Daily Papers

The project automatically fetches the latest papers from arXiv based on keywords.

The subheadings in the README file represent the search keywords.

Only the most recent articles for each keyword are retained, up to a maximum of 100 papers.

You can click the 'Watch' button to receive daily email notifications.

Last update: 2026-01-07

Vision Language Action

Title	Date	Abstract	Comment
Limited Linguistic Diversity in Embodied AI Datasets	2026-01-06	Show Language plays a critical role in Vision-Language-Action (VLA) models, yet the linguistic characteristics of the datasets used to train and evaluate these systems remain poorly documented. In this work, we present a systematic dataset audit of several widely used VLA corpora, aiming to characterize what kinds of instructions these datasets actually contain and how much linguistic variety they provide. We quantify instruction language along complementary dimensions-including lexical variety, duplication and overlap, semantic similarity, and syntactic complexity. Our analysis shows that many datasets rely on highly repetitive, template-like commands with limited structural variation, yielding a narrow distribution of instruction forms. We position these findings as descriptive documentation of the language signal available in current VLA training and evaluation data, intended to support more detailed dataset reporting, more principled dataset selection, and targeted curation or augmentation strategies that broaden language coverage.
SOP: A Scalable Online Post-Training System for Vision-Language-Action Models	2026-01-06	Show Vision-language-action (VLA) models achieve strong generalization through large-scale pre-training, but real-world deployment requires expert-level task proficiency in addition to broad generality. Existing post-training approaches for VLA models are typically offline, single-robot, or task-specific, limiting effective on-policy adaptation and scalable learning from real-world interaction. We introduce a Scalable Online Post-training (SOP) system that enables online, distributed, multi-task post-training of generalist VLA models directly in the physical world. SOP tightly couples execution and learning through a closed-loop architecture in which a fleet of robots continuously streams on-policy experience and human intervention signals to a centralized cloud learner, and asynchronously receives updated policies. This design supports prompt on-policy correction, scales experience collection through parallel deployment, and preserves generality during adaptation. SOP is agnostic to the choice of post-training algorithm; we instantiate it with both interactive imitation learning (HG-DAgger) and reinforcement learning (RECAP). Across a range of real-world manipulation tasks including cloth folding, box assembly, and grocery restocking, we show that SOP substantially improves the performance of large pretrained VLA models while maintaining a single shared policy across tasks. Effective post-training can be achieved within hours of real-world interaction, and performance scales near-linearly with the number of robots in the fleet. These results suggest that tightly coupling online learning with fleet-scale deployment is instrumental to enabling efficient, reliable, and scalable post-training of generalist robot policies in the physical world.
RoboMIND 2.0: A Multimodal, Bimanual Mobile Manipulation Dataset for Generalizable Embodied Intelligence	2026-01-06	Show While data-driven imitation learning has revolutionized robotic manipulation, current approaches remain constrained by the scarcity of large-scale, diverse real-world demonstrations. Consequently, the ability of existing models to generalize across long-horizon bimanual tasks and mobile manipulation in unstructured environments remains limited. To bridge this gap, we present RoboMIND 2.0, a comprehensive real-world dataset comprising over 310K dual-arm manipulation trajectories collected across six distinct robot embodiments and 739 complex tasks. Crucially, to support research in contact-rich and spatially extended tasks, the dataset incorporates 12K tactile-enhanced episodes and 20K mobile manipulation trajectories. Complementing this physical data, we construct high-fidelity digital twins of our real-world environments, releasing an additional 20K-trajectory simulated dataset to facilitate robust sim-to-real transfer. To fully exploit the potential of RoboMIND 2.0, we propose MIND-2 system, a hierarchical dual-system frame-work optimized via offline reinforcement learning. MIND-2 integrates a high-level semantic planner (MIND-2-VLM) to decompose abstract natural language instructions into grounded subgoals, coupled with a low-level Vision-Language-Action executor (MIND-2-VLA), which generates precise, proprioception-aware motor actions.
InternVLA-A1: Unifying Understanding, Generation and Action for Robotic Manipulation	2026-01-05	Show Prevalent Vision-Language-Action (VLA) models are typically built upon Multimodal Large Language Models (MLLMs) and demonstrate exceptional proficiency in semantic understanding, but they inherently lack the capability to deduce physical world dynamics. Consequently, recent approaches have shifted toward World Models, typically formulated via video prediction; however, these methods often suffer from a lack of semantic grounding and exhibit brittleness when handling prediction errors. To synergize semantic understanding with dynamic predictive capabilities, we present InternVLA-A1. This model employs a unified Mixture-of-Transformers architecture, coordinating three experts for scene understanding, visual foresight generation, and action execution. These components interact seamlessly through a unified masked self-attention mechanism. Building upon InternVL3 and Qwen3-VL, we instantiate InternVLA-A1 at 2B and 3B parameter scales. We pre-train these models on hybrid synthetic-real datasets spanning InternData-A1 and Agibot-World, covering over 533M frames. This hybrid training strategy effectively harnesses the diversity of synthetic simulation data while minimizing the sim-to-real gap. We evaluated InternVLA-A1 across 12 real-world robotic tasks and simulation benchmark. It significantly outperforms leading models like pi0 and GR00T N1.5, achieving a 14.5% improvement in daily tasks and a 40%-73.3% boost in dynamic settings, such as conveyor belt sorting.	Homep... Homepage: https://internrobotics.github.io/internvla-a1.github.io/
CycleVLA: Proactive Self-Correcting Vision-Language-Action Models via Subtask Backtracking and Minimum Bayes Risk Decoding	2026-01-05	Show Current work on robot failure detection and correction typically operate in a post hoc manner, analyzing errors and applying corrections only after failures occur. This work introduces CycleVLA, a system that equips Vision-Language-Action models (VLAs) with proactive self-correction, the capability to anticipate incipient failures and recover before they fully manifest during execution. CycleVLA achieves this by integrating a progress-aware VLA that flags critical subtask transition points where failures most frequently occur, a VLM-based failure predictor and planner that triggers subtask backtracking upon predicted failure, and a test-time scaling strategy based on Minimum Bayes Risk (MBR) decoding to improve retry success after backtracking. Extensive experiments show that CycleVLA improves performance for both well-trained and under-trained VLAs, and that MBR serves as an effective zero-shot test-time scaling strategy for VLAs. Project Page: https://dannymcy.github.io/cyclevla/	Proje... Project Page: https://dannymcy.github.io/cyclevla/
SurgWorld: Learning Surgical Robot Policies from Videos via World Modeling	2026-01-05	Show Data scarcity remains a fundamental barrier to achieving fully autonomous surgical robots. While large scale vision language action (VLA) models have shown impressive generalization in household and industrial manipulation by leveraging paired video action data from diverse domains, surgical robotics suffers from the paucity of datasets that include both visual observations and accurate robot kinematics. In contrast, vast corpora of surgical videos exist, but they lack corresponding action labels, preventing direct application of imitation learning or VLA training. In this work, we aim to alleviate this problem by learning policy models from SurgWorld, a world model designed for surgical physical AI. We curated the Surgical Action Text Alignment (SATA) dataset with detailed action description specifically for surgical robots. Then we built SurgeWorld based on the most advanced physical AI world model and SATA. It's able to generate diverse, generalizable and realistic surgery videos. We are also the first to use an inverse dynamics model to infer pseudokinematics from synthetic surgical videos, producing synthetic paired video action data. We demonstrate that a surgical VLA policy trained with these augmented data significantly outperforms models trained only on real demonstrations on a real surgical robot platform. Our approach offers a scalable path toward autonomous surgical skill acquisition by leveraging the abundance of unlabeled surgical video and generative world modeling, thus opening the door to generalizable and data efficient surgical robot policies.
Action-Sketcher: From Reasoning to Action via Visual Sketches for Long-Horizon Robotic Manipulation	2026-01-04	Show Long-horizon robotic manipulation is increasingly important for real-world deployment, requiring spatial disambiguation in complex layouts and temporal resilience under dynamic interaction. However, existing end-to-end and hierarchical Vision-Language-Action (VLA) policies often rely on text-only cues while keeping plan intent latent, which undermines referential grounding in cluttered or underspecified scenes, impedes effective task decomposition of long-horizon goals with close-loop interaction, and limits causal explanation by obscuring the rationale behind action choices. To address these issues, we first introduce Visual Sketch, an implausible visual intermediate that renders points, boxes, arrows, and typed relations in the robot's current views to externalize spatial intent, connect language to scene geometry. Building on Visual Sketch, we present Action-Sketcher, a VLA framework that operates in a cyclic See-Think-Sketch-Act workflow coordinated by adaptive token-gated strategy for reasoning triggers, sketch revision, and action issuance, thereby supporting reactive corrections and human interaction while preserving real-time action prediction. To enable scalable training and evaluation, we curate diverse corpus with interleaved images, text, Visual Sketch supervision, and action sequences, and train Action-Sketcher with a multi-stage curriculum recipe that combines interleaved sequence alignment for modality unification, language-to-sketch consistency for precise linguistic grounding, and imitation learning augmented with sketch-to-action reinforcement for robustness. Extensive experiments on cluttered scenes and multi-object tasks, in simulation and on real-world tasks, show improved long-horizon success, stronger robustness to dynamic scene changes, and enhanced interpretability via editable sketches and step-wise plans. Project website: https://action-sketcher.github.io	26 pages, 14 figures
Vision-Language-Action Models for Autonomous Driving: Past, Present, and Future	2026-01-04	Show Autonomous driving has long relied on modular "Perception-Decision-Action" pipelines, where hand-crafted interfaces and rule-based components often break down in complex or long-tailed scenarios. Their cascaded design further propagates perception errors, degrading downstream planning and control. Vision-Action (VA) models address some limitations by learning direct mappings from visual inputs to actions, but they remain opaque, sensitive to distribution shifts, and lack structured reasoning or instruction-following capabilities. Recent progress in Large Language Models (LLMs) and multimodal learning has motivated the emergence of Vision-Language-Action (VLA) frameworks, which integrate perception with language-grounded decision making. By unifying visual understanding, linguistic reasoning, and actionable outputs, VLAs offer a pathway toward more interpretable, generalizable, and human-aligned driving policies. This work provides a structured characterization of the emerging VLA landscape for autonomous driving. We trace the evolution from early VA approaches to modern VLA frameworks and organize existing methods into two principal paradigms: End-to-End VLA, which integrates perception, reasoning, and planning within a single model, and Dual-System VLA, which separates slow deliberation (via VLMs) from fast, safety-critical execution (via planners). Within these paradigms, we further distinguish subclasses such as textual vs. numerical action generators and explicit vs. implicit guidance mechanisms. We also summarize representative datasets and benchmarks for evaluating VLA-based driving systems and highlight key challenges and open directions, including robustness, interpretability, and instruction fidelity. Overall, this work aims to establish a coherent foundation for advancing human-compatible autonomous driving systems.	Surve... Survey; 47 pages, 7 figures, 9 tables; GitHub Repo at https://github.com/worldbench/awesome-vla-for-ad
Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbone	2026-01-04	Show While autoregressive Large Vision-Language Models (VLMs) have achieved remarkable success, their sequential generation often limits their efficacy in complex visual planning and dynamic robotic control. In this work, we investigate the potential of constructing Vision-Language Models upon diffusion-based large language models (dLLMs) to overcome these limitations. We introduce Dream-VL, an open diffusion-based VLM (dVLM) that achieves state-of-the-art performance among previous dVLMs. Dream-VL is comparable to top-tier AR-based VLMs trained on open data on various benchmarks but exhibits superior potential when applied to visual planning tasks. Building upon Dream-VL, we introduce Dream-VLA, a dLLM-based Vision-Language-Action model (dVLA) developed through continuous pre-training on open robotic datasets. We demonstrate that the natively bidirectional nature of this diffusion backbone serves as a superior foundation for VLA tasks, inherently suited for action chunking and parallel generation, leading to significantly faster convergence in downstream fine-tuning. Dream-VLA achieves top-tier performance of 97.2% average success rate on LIBERO, 71.4% overall average on SimplerEnv-Bridge, and 60.5% overall average on SimplerEnv-Fractal, surpassing leading models such as $π_0$ and GR00T-N1. We also validate that dVLMs surpass AR baselines on downstream tasks across different training objectives. We release both Dream-VL and Dream-VLA to facilitate further research in the community.	Add r... Add real-world experiments
Value Vision-Language-Action Planning & Search	2026-01-02	Show Vision-Language-Action (VLA) models have emerged as powerful generalist policies for robotic manipulation, yet they remain fundamentally limited by their reliance on behavior cloning, leading to brittleness under distribution shift. While augmenting pretrained models with test-time search algorithms like Monte Carlo Tree Search (MCTS) can mitigate these failures, existing formulations rely solely on the VLA prior for guidance, lacking a grounded estimate of expected future return. Consequently, when the prior is inaccurate, the planner can only correct action selection via the exploration term, which requires extensive simulation to become effective. To address this limitation, we introduce Value Vision-Language-Action Planning and Search (V-VLAPS), a framework that augments MCTS with a lightweight, learnable value function. By training a simple multilayer perceptron (MLP) on the latent representations of a fixed VLA backbone (Octo), we provide the search with an explicit success signal that biases action selection toward high-value regions. We evaluate V-VLAPS on the LIBERO robotic manipulation suite, demonstrating that our value-guided search improves success rates by over 5 percentage points while reducing the average number of MCTS simulations by 5-15 percent compared to baselines that rely only on the VLA prior.	10 pages, 3 figures
Unified Embodied VLM Reasoning with Robotic Action via Autoregressive Discretized Pre-training	2026-01-01	Show General-purpose robotic systems operating in open-world environments must achieve both broad generalization and high-precision action execution, a combination that remains challenging for existing Vision-Language-Action (VLA) models. While large Vision-Language Models (VLMs) improve semantic generalization, insufficient embodied reasoning leads to brittle behavior, and conversely, strong reasoning alone is inadequate without precise control. To provide a decoupled and quantitative assessment of this bottleneck, we introduce Embodied Reasoning Intelligence Quotient (ERIQ), a large-scale embodied reasoning benchmark in robotic manipulation, comprising 6K+ question-answer pairs across four reasoning dimensions. By decoupling reasoning from execution, ERIQ enables systematic evaluation and reveals a strong positive correlation between embodied reasoning capability and end-to-end VLA generalization. To bridge the gap from reasoning to precise execution, we propose FACT, a flow-matching-based action tokenizer that converts continuous control into discrete sequences while preserving high-fidelity trajectory reconstruction. The resulting GenieReasoner jointly optimizes reasoning and action in a unified space, outperforming both continuous-action and prior discrete-action baselines in real-world tasks. Together, ERIQ and FACT provide a principled framework for diagnosing and overcoming the reasoning-precision trade-off, advancing robust, general-purpose robotic manipulation. Project page: https://geniereasoner.github.io/GenieReasoner/
Dichotomous Diffusion Policy Optimization	2025-12-31	Show Diffusion-based policies have gained growing popularity in solving a wide range of decision-making tasks due to their superior expressiveness and controllable generation during inference. However, effectively training large diffusion policies using reinforcement learning (RL) remains challenging. Existing methods either suffer from unstable training due to directly maximizing value objectives, or face computational issues due to relying on crude Gaussian likelihood approximation, which requires a large amount of sufficiently small denoising steps. In this work, we propose DIPOLE (Dichotomous diffusion Policy improvement), a novel RL algorithm designed for stable and controllable diffusion policy optimization. We begin by revisiting the KL-regularized objective in RL, which offers a desirable weighted regression objective for diffusion policy extraction, but often struggles to balance greediness and stability. We then formulate a greedified policy regularization scheme, which naturally enables decomposing the optimal policy into a pair of stably learned dichotomous policies: one aims at reward maximization, and the other focuses on reward minimization. Under such a design, optimized actions can be generated by linearly combining the scores of dichotomous policies during inference, thereby enabling flexible control over the level of greediness.Evaluations in offline and offline-to-online RL settings on ExORL and OGBench demonstrate the effectiveness of our approach. We also use DIPOLE to train a large vision-language-action (VLA) model for end-to-end autonomous driving (AD) and evaluate it on the large-scale real-world AD benchmark NAVSIM, highlighting its potential for complex real-world applications.
ColaVLA: Leveraging Cognitive Latent Reasoning for Hierarchical Parallel Trajectory Planning in Autonomous Driving	2025-12-31	Show Autonomous driving requires generating safe and reliable trajectories from complex multimodal inputs. Traditional modular pipelines separate perception, prediction, and planning, while recent end-to-end (E2E) systems learn them jointly. Vision-language models (VLMs) further enrich this paradigm by introducing cross-modal priors and commonsense reasoning, yet current VLM-based planners face three key challenges: (i) a mismatch between discrete text reasoning and continuous control, (ii) high latency from autoregressive chain-of-thought decoding, and (iii) inefficient or non-causal planners that limit real-time deployment. We propose ColaVLA, a unified vision-language-action framework that transfers reasoning from text to a unified latent space and couples it with a hierarchical, parallel trajectory decoder. The Cognitive Latent Reasoner compresses scene understanding into compact, decision-oriented meta-action embeddings through ego-adaptive selection and only two VLM forward passes. The Hierarchical Parallel Planner then generates multi-scale, causality-consistent trajectories in a single forward pass. Together, these components preserve the generalization and interpretability of VLMs while enabling efficient, accurate and safe trajectory generation. Experiments on the nuScenes benchmark show that ColaVLA achieves state-of-the-art performance in both open-loop and closed-loop settings with favorable efficiency and robustness.	11 pa... 11 pages, 4 figures. Project page: https://pqh22.github.io/projects/ColaVLA/index.html
VLA-RAIL: A Real-Time Asynchronous Inference Linker for VLA Models and Robots	2025-12-31	Show Vision-Language-Action (VLA) models have achieved remarkable breakthroughs in robotics, with the action chunk playing a dominant role in these advances. Given the real-time and continuous nature of robotic motion control, the strategies for fusing a queue of successive action chunks have a profound impact on the overall performance of VLA models. Existing methods suffer from jitter, stalling, or even pauses in robotic action execution, which not only limits the achievable execution speed but also reduces the overall success rate of task completion. This paper introduces VLA-RAIL (A Real-Time Asynchronous Inference Linker), a novel framework designed to address these issues by conducting model inference and robot motion control asynchronously and guaranteeing smooth, continuous, and high-speed action execution. The core contributions of the paper are two fold: a Trajectory Smoother that effectively filters out the noise and jitter in the trajectory of one action chunk using polynomial fitting and a Chunk Fuser that seamlessly align the current executing trajectory and the newly arrived chunk, ensuring position, velocity, and acceleration continuity between two successive action chunks. We validate the effectiveness of VLA-RAIL on a benchmark of dynamic simulation tasks and several real-world manipulation tasks. Experimental results demonstrate that VLA-RAIL significantly reduces motion jitter, enhances execution speed, and improves task success rates, which will become a key infrastructure for the large-scale deployment of VLA models.
Counterfactual VLA: Self-Reflective Vision-Language-Action Model with Adaptive Reasoning	2025-12-30	Show Recent reasoning-augmented Vision-Language-Action (VLA) models have improved the interpretability of end-to-end autonomous driving by generating intermediate reasoning traces. Yet these models primarily describe what they perceive and intend to do, rarely questioning whether their planned actions are safe or appropriate. This work introduces Counterfactual VLA (CF-VLA), a self-reflective VLA framework that enables the model to reason about and revise its planned actions before execution. CF-VLA first generates time-segmented meta-actions that summarize driving intent, and then performs counterfactual reasoning conditioned on both the meta-actions and the visual context. This step simulates potential outcomes, identifies unsafe behaviors, and outputs corrected meta-actions that guide the final trajectory generation. To efficiently obtain such self-reflective capabilities, we propose a rollout-filter-label pipeline that mines high-value scenes from a base (non-counterfactual) VLA's rollouts and labels counterfactual reasoning traces for subsequent training rounds. Experiments on large-scale driving datasets show that CF-VLA improves trajectory accuracy by up to 17.6%, enhances safety metrics by 20.5%, and exhibits adaptive thinking: it only enables counterfactual reasoning in challenging scenarios. By transforming reasoning traces from one-shot descriptions to causal self-correction signals, CF-VLA takes a step toward self-reflective autonomous driving agents that learn to think before they act.
GR-Dexter Technical Report	2025-12-30	Show Vision-language-action (VLA) models have enabled language-conditioned, long-horizon robot manipulation, but most existing systems are limited to grippers. Scaling VLA policies to bimanual robots with high degree-of-freedom (DoF) dexterous hands remains challenging due to the expanded action space, frequent hand-object occlusions, and the cost of collecting real-robot data. We present GR-Dexter, a holistic hardware-model-data framework for VLA-based generalist manipulation on a bimanual dexterous-hand robot. Our approach combines the design of a compact 21-DoF robotic hand, an intuitive bimanual teleoperation system for real-robot data collection, and a training recipe that leverages teleoperated robot trajectories together with large-scale vision-language and carefully curated cross-embodiment datasets. Across real-world evaluations spanning long-horizon everyday manipulation and generalizable pick-and-place, GR-Dexter achieves strong in-domain performance and improved robustness to unseen objects and unseen instructions. We hope GR-Dexter serves as a practical step toward generalist dexterous-hand robotic manipulation.
OTTER: A Vision-Language-Action Model with Text-Aware Visual Feature Extraction	2025-12-30	Show Vision-Language-Action (VLA) models aim to predict robotic actions based on visual observations and language instructions. Existing approaches require fine-tuning pre-trained visionlanguage models (VLMs) as visual and language features are independently fed into downstream policies, degrading the pre-trained semantic alignments. We propose OTTER, a novel VLA architecture that leverages these existing alignments through explicit, text-aware visual feature extraction. Instead of processing all visual features, OTTER selectively extracts and passes only task-relevant visual features that are semantically aligned with the language instruction to the policy transformer. This allows OTTER to keep the pre-trained vision-language encoders frozen. Thereby, OTTER preserves and utilizes the rich semantic understanding learned from large-scale pre-training, enabling strong zero-shot generalization capabilities. In simulation and real-world experiments, OTTER significantly outperforms existing VLA models, demonstrating strong zeroshot generalization to novel objects and environments. Video, code, checkpoints, and dataset: https://ottervla.github.io/.
Learning to Feel the Future: DreamTacVLA for Contact-Rich Manipulation	2025-12-29	Show Vision-Language-Action (VLA) models have shown remarkable generalization by mapping web-scale knowledge to robotic control, yet they remain blind to physical contact. Consequently, they struggle with contact-rich manipulation tasks that require reasoning about force, texture, and slip. While some approaches incorporate low-dimensional tactile signals, they fail to capture the high-resolution dynamics essential for such interactions. To address this limitation, we introduce DreamTacVLA, a framework that grounds VLA models in contact physics by learning to feel the future. Our model adopts a hierarchical perception scheme in which high-resolution tactile images serve as micro-vision inputs coupled with wrist-camera local vision and third-person macro vision. To reconcile these multi-scale sensory streams, we first train a unified policy with a Hierarchical Spatial Alignment (HSA) loss that aligns tactile tokens with their spatial counterparts in the wrist and third-person views. To further deepen the model's understanding of fine-grained contact dynamics, we finetune the system with a tactile world model that predicts future tactile signals. To mitigate tactile data scarcity and the wear-prone nature of tactile sensors, we construct a hybrid large-scale dataset sourced from both high-fidelity digital twin and real-world experiments. By anticipating upcoming tactile states, DreamTacVLA acquires a rich model of contact physics and conditions its actions on both real observations and imagined consequences. Across contact-rich manipulation tasks, it outperforms state-of-the-art VLA baselines, achieving up to 95% success, highlighting the importance of understanding physical contact for robust, touch-aware robotic agents.
VLA-Arena: An Open-Source Framework for Benchmarking Vision-Language-Action Models	2025-12-27	Show While Vision-Language-Action models (VLAs) are rapidly advancing towards generalist robot policies, it remains difficult to quantitatively understand their limits and failure modes. To address this, we introduce a comprehensive benchmark called VLA-Arena. We propose a novel structured task design framework to quantify difficulty across three orthogonal axes: (1) Task Structure, (2) Language Command, and (3) Visual Observation. This allows us to systematically design tasks with fine-grained difficulty levels, enabling a precise measurement of model capability frontiers. For Task Structure, VLA-Arena's 170 tasks are grouped into four dimensions: Safety, Distractor, Extrapolation, and Long Horizon. Each task is designed with three difficulty levels (L0-L2), with fine-tuning performed exclusively on L0 to assess general capability. Orthogonal to this, language (W0-W4) and visual (V0-V4) perturbations can be applied to any task to enable a decoupled analysis of robustness. Our extensive evaluation of state-of-the-art VLAs reveals several critical limitations, including a strong tendency toward memorization over generalization, asymmetric robustness, a lack of consideration for safety constraints, and an inability to compose learned skills for long-horizon tasks. To foster research addressing these challenges and ensure reproducibility, we provide the complete VLA-Arena framework, including an end-to-end toolchain from task definition to automated evaluation and the VLA-Arena-S/M/L datasets for fine-tuning. Our benchmark, data, models, and leaderboard are available at https://vla-arena.github.io.
Clutter-Resistant Vision-Language-Action Models through Object-Centric and Geometry Grounding	2025-12-27	Show Recent Vision-Language-Action (VLA) models have made impressive progress toward general-purpose robotic manipulation by post-training large Vision-Language Models (VLMs) for action prediction. Yet most VLAs entangle perception and control in a monolithic pipeline optimized purely for action, which can erode language-conditioned grounding. In our real-world tabletop tests, policies over-grasp when the target is absent, are distracted by clutter, and overfit to background appearance. To address these issues, we propose OBEYED-VLA (OBject-centric and gEometrY groundED VLA), a framework that explicitly disentangles perceptual grounding from action reasoning. Instead of operating directly on raw RGB, OBEYED-VLA augments VLAs with a perception module that grounds multi-view inputs into task-conditioned, object-centric, and geometry-aware observations. This module includes a VLM-based object-centric grounding stage that selects task-relevant object regions across camera views, along with a complementary geometric grounding stage that emphasizes the 3D structure of these objects over their appearance. The resulting grounded views are then fed to a pretrained VLA policy, which we fine-tune exclusively on single-object demonstrations collected without environmental clutter or non-target objects. On a real-world UR10e tabletop setup, OBEYED-VLA substantially improves robustness over strong VLA baselines across four challenging regimes and multiple difficulty levels: distractor objects, absent-target rejection, background appearance changes, and cluttered manipulation of unseen objects. Ablation studies confirm that both semantic grounding and geometry-aware grounding are critical to these gains. Overall, the results indicate that making perception an explicit, object-centric component is an effective way to strengthen and generalize VLA-based robotic manipulation.	Under... Under review. Project website: https://uark-aicv.github.io/OBEYED_VLA
Emergence of Human to Robot Transfer in Vision-Language-Action Models	2025-12-27	Show Vision-language-action (VLA) models can enable broad open world generalization, but require large and diverse datasets. It is appealing to consider whether some of this data can come from human videos, which cover diverse real-world situations and are easy to obtain. However, it is difficult to train VLAs with human videos alone, and establishing a mapping between humans and robots requires manual engineering and presents a major research challenge. Drawing inspiration from advances in large language models, where the ability to learn from diverse supervision emerges with scale, we ask whether a similar phenomenon holds for VLAs that incorporate human video data. We introduce a simple co-training recipe, and find that human-to-robot transfer emerges once the VLA is pre-trained on sufficient scenes, tasks, and embodiments. Our analysis suggests that this emergent capability arises because diverse pretraining produces embodiment-agnostic representations for human and robot data. We validate these findings through a series of experiments probing human to robot skill transfer and find that with sufficiently diverse robot pre-training our method can nearly double the performance on generalization settings seen only in human data.
LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models	2025-12-26	Show Visual-Language-Action (VLA) models report impressive success rates on robotic manipulation benchmarks, yet these results may mask fundamental weaknesses in robustness. We perform a systematic vulnerability analysis by introducing controlled perturbations across seven dimensions: objects layout, camera viewpoints, robot initial states, language instructions, light conditions, background textures and sensor noise. We comprehensively analyzed multiple state-of-the-art models and revealed consistent brittleness beneath apparent competence. Our analysis exposes critical weaknesses: models exhibit extreme sensitivity to perturbation factors, including camera viewpoints and robot initial states, with performance dropping from 95% to below 30% under modest perturbations. Surprisingly, models are largely insensitive to language variations, with further experiments revealing that models tend to ignore language instructions completely. Our findings challenge the assumption that high benchmark scores equate to true competency and highlight the need for evaluation practices that assess reliability under realistic variation.
StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision	2025-12-26	Show Stereo cameras closely mimic human binocular vision, providing rich spatial cues critical for precise robotic manipulation. Despite their advantage, the adoption of stereo vision in vision-language-action models (VLAs) remains underexplored. In this work, we present StereoVLA, a VLA model that leverages rich geometric cues from stereo vision. We propose a novel Geometric-Semantic Feature Extraction module that utilizes vision foundation models to extract and fuse two key features: 1) geometric features from subtle stereo-view differences for spatial perception; 2) semantic-rich features from the monocular view for instruction following. Additionally, we propose an auxiliary Interaction-Region Depth Estimation task to further enhance spatial perception and accelerate model convergence. Extensive experiments show that our approach outperforms baselines by a large margin in diverse tasks under the stereo setting and demonstrates strong robustness to camera pose variations.
Motus: A Unified Latent Action World Model	2025-12-25	Show While a general embodied agent must function as a unified system, current methods are built on isolated models for understanding, world modeling, and control. This fragmentation prevents unifying multimodal generative capabilities and hinders learning from large-scale, heterogeneous data. In this paper, we propose Motus, a unified latent action world model that leverages existing general pretrained models and rich, sharable motion information. Motus introduces a Mixture-of-Transformer (MoT) architecture to integrate three experts (i.e., understanding, video generation, and action) and adopts a UniDiffuser-style scheduler to enable flexible switching between different modeling modes (i.e., world models, vision-language-action models, inverse dynamics models, video generation models, and video-action joint prediction models). Motus further leverages the optical flow to learn latent actions and adopts a recipe with three-phase training pipeline and six-layer data pyramid, thereby extracting pixel-level "delta action" and enabling large-scale action pretraining. Experiments show that Motus achieves superior performance against state-of-the-art methods in both simulation (a +15% improvement over X-VLA and a +45% improvement over Pi0.5) and real-world scenarios(improved by +11~48%), demonstrating unified modeling of all functionalities and priors significantly benefits downstream robotic tasks.
STARE-VLA: Progressive Stage-Aware Reinforcement for Fine-Tuning Vision-Language-Action Models	2025-12-23	Show Recent advances in Vision-Language-Action (VLA) models, powered by large language models and reinforcement learning-based fine-tuning, have shown remarkable progress in robotic manipulation. Existing methods often treat long-horizon actions as linguistic sequences and apply trajectory-level optimization methods such as Trajectory-wise Preference Optimization (TPO) or Proximal Policy Optimization (PPO), leading to coarse credit assignment and unstable training. However, unlike language, where a unified semantic meaning is preserved despite flexible sentence order, action trajectories progress through causally chained stages with different learning difficulties. This motivates progressive stage optimization. Thereby, we present Stage-Aware Reinforcement (STARE), a module that decomposes a long-horizon action trajectory into semantically meaningful stages and provides dense, interpretable, and stage-aligned reinforcement signals. Integrating STARE into TPO and PPO, we yield Stage-Aware TPO (STA-TPO) and Stage-Aware PPO (STA-PPO) for offline stage-wise preference and online intra-stage interaction, respectively. Further building on supervised fine-tuning as initialization, we propose the Imitation -> Preference -> Interaction (IPI), a serial fine-tuning pipeline for improving action accuracy in VLA models. Experiments on SimplerEnv and ManiSkill3 demonstrate substantial gains, achieving state-of-the-art success rates of 98.0 percent on SimplerEnv and 96.4 percent on ManiSkill3 tasks.
Reinforcement Learning for Large Model: A Survey	2025-12-23	Show Recent advances at the intersection of reinforcement learning (RL) and visual intelligence have enabled agents that not only perceive complex visual scenes but also reason, generate, and act within them. This survey offers a critical and up-to-date synthesis of the field. We first formalize visual RL problems and trace the evolution of policy-optimization strategies from RLHF to verifiable reward paradigms, and from Proximal Policy Optimization to Group Relative Policy Optimization. We then organize more than 200 representative works into four thematic pillars: multi-modal large language models, visual generation, unified model frameworks, and vision-language-action models. For each pillar we examine algorithmic design, reward engineering, benchmark progress, and we distill trends such as curriculum-driven training, preference-aligned diffusion, and unified reward modeling. Finally, we review evaluation protocols spanning set-level fidelity, sample-level preference, and state-level stability, and we identify open challenges that include sample efficiency, generalization, and safe deployment. Our goal is to provide researchers and practitioners with a coherent map of the rapidly expanding landscape of visual RL and to highlight promising directions for future inquiry. Resources are available at: https://github.com/weijiawu/Awesome-Visual-Reinforcement-Learning.	22 pages
ActionFlow: A Pipelined Action Acceleration for Vision Language Models on Edge	2025-12-23	Show Vision-Language-Action (VLA) models have emerged as a unified paradigm for robotic perception and control, enabling emergent generalization and long-horizon task execution. However, their deployment in dynamic, real-world environments is severely hin dered by high inference latency. While smooth robotic interaction requires control frequencies of 20 to 30 Hz, current VLA models typi cally operate at only 3-5 Hz on edge devices due to the memory bound nature of autoregressive decoding. Existing optimizations often require extensive retraining or compromise model accuracy. To bridge this gap, we introduce ActionFlow, a system-level inference framework tailored for resource-constrained edge plat forms. At the core of ActionFlow is a Cross-Request Pipelin ing strategy, a novel scheduler that redefines VLA inference as a macro-pipeline of micro-requests. The strategy intelligently batches memory-bound Decode phases with compute-bound Prefill phases across continuous time steps to maximize hardware utilization. Furthermore, to support this scheduling, we propose a Cross Request State Packed Forward operator and a Unified KV Ring Buffer, which fuse fragmented memory operations into efficient dense computations. Experimental results demonstrate that ActionFlow achieves a 2.55x improvement in FPS on the OpenVLA-7B model without retraining, enabling real-time dy namic manipulation on edge hardware. Our work is available at https://anonymous.4open.science/r/ActionFlow-1D47.
Continuous Vision-Language-Action Co-Learning with Semantic-Physical Alignment for Behavioral Cloning	2025-12-23	Show Language-conditioned manipulation facilitates human-robot interaction via behavioral cloning (BC), which learns control policies from human demonstrations and serves as a cornerstone of embodied AI. Overcoming compounding errors in sequential action decisions remains a central challenge to improving BC performance. Existing approaches mitigate compounding errors through data augmentation, expressive representation, or temporal abstraction. However, they suffer from physical discontinuities and semantic-physical misalignment, leading to inaccurate action cloning and intermittent execution. In this paper, we present Continuous vision-language-action Co-Learning with Semantic-Physical Alignment (CCoL), a novel BC framework that ensures temporally consistent execution and fine-grained semantic grounding. It generates robust and smooth action execution trajectories through continuous co-learning across vision, language, and proprioceptive inputs (e.g., robot internal states). Meanwhile, we anchor language semantics to visuomotor representations by a bidirectional cross-attention to learn contextual information for action generation, successfully overcoming the problem of semantic-physical misalignment. Extensive experiments show that CCoL achieves an average 8.0% relative improvement across three simulation suites, with up to 19.2% relative gain in human-demonstrated bimanual insertion tasks. Real-world tests on a 7-DoF robot further confirm CCoL's generalization under unseen and noisy object states.	Accep... Accepted at AAAI 2026, the Project website is available at https://qhemu.github.io/CCoL/
Asynchronous Fast-Slow Vision-Language-Action Policies for Whole-Body Robotic Manipulation	2025-12-23	Show Most Vision-Language-Action (VLA) systems integrate a Vision-Language Model (VLM) for semantic reasoning with an action expert generating continuous action signals, yet both typically run at a single unified frequency. As a result, policy performance is constrained by the low inference speed of large VLMs. This mandatory synchronous execution severely limits control stability and real-time performance in whole-body robotic manipulation, which involves more joints, larger motion spaces, and dynamically changing views. We introduce a truly asynchronous Fast-Slow VLA framework (DuoCore-FS), organizing the system into a fast pathway for high-frequency action generation and a slow pathway for rich VLM reasoning. The system is characterized by two key features. First, a latent representation buffer bridges the slow and fast systems. It stores instruction semantics and action-reasoning representation aligned with the scene-instruction context, providing high-level guidance to the fast pathway. Second, a whole-body action tokenizer provides a compact, unified representation of whole-body actions. Importantly, the VLM and action expert are still jointly trained end-to-end, preserving unified policy learning while enabling asynchronous execution. DuoCore-FS supports a 3B-parameter VLM while achieving 30 Hz whole-body action-chunk generation, approximately three times as fast as prior VLA models with comparable model sizes. Real-world whole-body manipulation experiments demonstrate improved task success rates and significantly enhanced responsiveness compared to synchronous Fast-Slow VLA baselines. The implementation of DuoCore-FS, including training, inference, and deployment, is provided to commercial users by Astribot as part of the Astribot robotic platform.
LoLA: Long Horizon Latent Action Learning for General Robot Manipulation	2025-12-23	Show The capability of performing long-horizon, language-guided robotic manipulation tasks critically relies on leveraging historical information and generating coherent action sequences. However, such capabilities are often overlooked by existing Vision-Language-Action (VLA) models. To solve this challenge, we propose LoLA (Long Horizon Latent Action Learning), a framework designed for robot manipulation that integrates long-term multi-view observations and robot proprioception to enable multi-step reasoning and action generation. We first employ Vision-Language Models to encode rich contextual features from historical sequences and multi-view observations. We further introduces a key module, State-Aware Latent Re-representation, which transforms visual inputs and language commands into actionable robot motion space. Unlike existing VLA approaches that merely concatenate robot proprioception (e.g., joint angles) with VL embeddings, this module leverages such robot states to explicitly ground VL representations in physical scale through a learnable "embodiment-anchored" latent space. We trained LoLA on diverse robotic pre-training datasets and conducted extensive evaluations on simulation benchmarks (SIMPLER and LIBERO), as well as two real-world tasks on Franka and Bi-Manual Aloha robots. Results show that LoLA significantly outperforms prior state-of-the-art methods (e.g., pi0), particularly in long-horizon manipulation tasks.
Bring My Cup! Personalizing Vision-Language-Action Models with Visual Attentive Prompting	2025-12-23	Show While Vision-Language-Action (VLA) models generalize well to generic instructions, they struggle with personalized commands such as "bring my cup", where the robot must act on one specific instance among visually similar objects. We study this setting of manipulating personal objects, in which a VLA must identify and control a user-specific object unseen during training using only a few reference images. To address this challenge, we propose Visual Attentive Prompting (VAP), a simple-yet-effective training-free perceptual adapter that equips frozen VLAs with top-down selective attention. VAP treats the reference images as a non-parametric visual memory, grounds the personal object in the scene through open-vocabulary detection and embedding-based matching, and then injects this grounding as a visual prompt by highlighting the object and rewriting the instruction. We construct two simulation benchmarks, Personalized-SIMPLER and Personalized-VLABench, and a real-world tabletop benchmark to evaluate personalized manipulation across multiple robots and tasks. Experiments show that VAP consistently outperforms generic policies and token-learning baselines in both success rate and correct-object manipulation, helping to bridge the gap between semantic understanding and instance-level control.
GR-RL: Going Dexterous and Precise for Long-Horizon Robotic Manipulation	2025-12-23	Show We present GR-RL, a robotic learning framework that turns a generalist vision-language-action (VLA) policy into a highly capable specialist for long-horizon dexterous manipulation. Assuming the optimality of human demonstrations is core to existing VLA policies. However, we claim that in highly dexterous and precise manipulation tasks, human demonstrations are noisy and suboptimal. GR-RL proposes a multi-stage training pipeline that filters, augments, and reinforces the demonstrations by reinforcement learning. First, GR-RL learns a vision-language-conditioned task progress, filters the demonstration trajectories, and only keeps the transitions that contribute positively to the progress. Specifically, we show that by directly applying offline RL with sparse reward, the resulting $Q$-values can be treated as a robust progress function. Next, we introduce morphological symmetry augmentation that greatly improves the generalization and performance of GR-RL. Lastly, to better align the VLA policy with its deployment behaviors for high-precision control, we perform online RL by learning a latent space noise predictor. With this pipeline, GR-RL is, to our knowledge, the first learning-based policy that can autonomously lace up a shoe by threading shoelaces through multiple eyelets with an 83.3% success rate, a task requiring long-horizon reasoning, millimeter-level precision, and compliant soft-body interaction. We hope GR-RL provides a step toward enabling generalist robot foundation models to specialize into reliable real-world experts.
Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policies	2025-12-22	Show Vision-Language-Action (VLA) models adapt large vision-language backbones to map images and instructions into robot actions. However, prevailing VLAs either generate actions auto-regressively in a fixed left-to-right order or attach separate MLP or diffusion heads outside the backbone, leading to fragmented information pathways and specialized training requirements that hinder a unified, scalable architecture. We present Discrete Diffusion VLA, a unified-transformer policy that models discretized action chunks with discrete diffusion. The design retains diffusion's progressive refinement paradigm while remaining natively compatible with the discrete token interface of VLMs. Our method achieves an adaptive decoding order that resolves easy action elements before harder ones and uses secondary re-masking to revisit uncertain predictions across refinement rounds, which improves consistency and enables robust error correction. This unified decoder preserves pre-trained vision-language priors, supports parallel decoding, breaks the autoregressive bottleneck, and reduces the number of function evaluations. Discrete Diffusion VLA achieves 96.3% avg. success rates on LIBERO, 71.2% visual matching on SimplerEnv-Fractal and 54.2% overall on SimplerEnv-Bridge. We also provide ablation study on vision-language ability retention on LIBERO-OOD (Out-of-Distribution) benchmark, with our method improving over autoregressive, MLP decoder and continuous diffusion baselines. These findings indicate that discrete-diffusion VLA supports precise action modeling and consistent training, laying groundwork for scaling VLA to larger models and datasets. Our code is available at https://github.com/Liang-ZX/DiscreteDiffusionVLA/tree/libero.	New e... New experiments on VL retention and new ablations. 18 pages
REALM: A Real-to-Sim Validated Benchmark for Generalization in Robotic Manipulation	2025-12-22	Show Vision-Language-Action (VLA) models empower robots to understand and execute tasks described by natural language instructions. However, a key challenge lies in their ability to generalize beyond the specific environments and conditions they were trained on, which is presently difficult and expensive to evaluate in the real-world. To address this gap, we present REALM, a new simulation environment and benchmark designed to evaluate the generalization capabilities of VLA models, with a specific emphasis on establishing a strong correlation between simulated and real-world performance through high-fidelity visuals and aligned robot control. Our environment offers a suite of 15 perturbation factors, 7 manipulation skills, and more than 3,500 objects. Finally, we establish two task sets that form our benchmark and evaluate the π_{0}, π_{0}-FAST, and GR00T N1.5 VLA models, showing that generalization and robustness remain an open challenge. More broadly, we also show that simulation gives us a valuable proxy for the real-world and allows us to systematically probe for and quantify the weaknesses and failure modes of VLAs. Project page: https://martin-sedlacek.com/realm	9 pages, 10 figures
Confidence Calibration in Vision-Language-Action Models	2025-12-22	Show Trustworthy robot behavior requires not only high levels of task success but also that the robot can reliably quantify how likely it is to succeed. To this end, we present a first-of-its-kind study of confidence calibration in vision-language-action (VLA) foundation models, which map visual observations and natural language instructions to low-level robot motor commands. We establish a confidence baseline for VLAs, examine how task success relates to calibration error and how calibration evolves over time, and introduce two lightweight techniques to remedy the miscalibration we observe: prompt ensembles and action-wise Platt scaling. Our aim in this study is to begin to develop the tools and conceptual understanding necessary to render VLAs both highly performant and highly trustworthy via reliable uncertainty quantification.	38 pa... 38 pages, 19 figures; additional experiments with VLA variants
Open-Source Multimodal Moxin Models with Moxin-VLM and Moxin-VLA	2025-12-22	Show Recently, Large Language Models (LLMs) have undergone a significant transformation, marked by a rapid rise in both their popularity and capabilities. Leading this evolution are proprietary LLMs like GPT-4 and GPT-o1, which have captured widespread attention in the AI community due to their remarkable performance and versatility. Simultaneously, open-source LLMs, such as LLaMA and Mistral, have made great contributions to the ever-increasing popularity of LLMs due to the ease to customize and deploy the models across diverse applications. Moxin 7B is introduced as a fully open-source LLM developed in accordance with the Model Openness Framework, which moves beyond the simple sharing of model weights to embrace complete transparency in training, datasets, and implementation detail, thus fostering a more inclusive and collaborative research environment that can sustain a healthy open-source ecosystem. To further equip Moxin with various capabilities in different tasks, we develop three variants based on Moxin, including Moxin-VLM, Moxin-VLA, and Moxin-Chinese, which target the vision-language, vision-language-action, and Chinese capabilities, respectively. Experiments show that our models achieve superior performance in various evaluations. We adopt open-source framework and open data for the training. We release our models, along with the available data and code to derive these models.
QDepth-VLA: Quantized Depth Prediction as Auxiliary Supervision for Vision-Language-Action Models	2025-12-22	Show Spatial perception and reasoning are crucial for Vision-Language-Action (VLA) models to accomplish fine-grained manipulation tasks. However, existing approaches often lack the ability to understand and reason over the essential 3D structures necessary for precise control. To address this limitation, we propose QDepth-VLA, a general framework that augments VLA models with an auxiliary depth prediction task. A dedicated depth expert is designed to predict quantized latent tokens of depth maps obtained from a VQ-VAE encoder, enabling the model to learn depth-aware representations that capture critical geometric cues. Experimental results on the simulation benchmarks and real-world tasks demonstrate that QDepth-VLA yields strong spatial reasoning and competitive performance on manipulation tasks.
Point What You Mean: Visually Grounded Instruction Policy	2025-12-22	Show Vision-Language-Action (VLA) models align vision and language with embodied control, but their object referring ability remains limited when relying solely on text prompt, especially in cluttered or out-of-distribution (OOD) scenes. In this study, we introduce the Point-VLA, a plug-and-play policy that augments language instructions with explicit visual cues (e.g., bounding boxes) to resolve referential ambiguity and enable precise object-level grounding. To efficiently scale visually grounded datasets, we further develop an automatic data annotation pipeline requiring minimal human effort. We evaluate Point-VLA on diverse real-world referring tasks and observe consistently stronger performance than text-only instruction VLAs, particularly in cluttered or unseen-object scenarios, with robust generalization. These results demonstrate that Point-VLA effectively resolves object referring ambiguity through pixel-level visual grounding, achieving more generalizable embodied control.
Task adaptation of Vision-Language-Action model: 1st Place Solution for the 2025 BEHAVIOR Challenge	2025-12-21	Show We present a vision-action policy that won 1st place in the 2025 BEHAVIOR Challenge - a large-scale benchmark featuring 50 diverse long-horizon household tasks in photo-realistic simulation, requiring bimanual manipulation, navigation, and context-aware decision making. Building on the Pi0.5 architecture, we introduce several innovations. Our primary contribution is correlated noise for flow matching, which improves training efficiency and enables correlation-aware inpainting for smooth action sequences. We also apply learnable mixed-layer attention and System 2 stage tracking for ambiguity resolution. Training employs multi-sample flow matching to reduce variance, while inference uses action compression and challenge-specific correction rules. Our approach achieves 26% q-score across all 50 tasks on both public and private leaderboards.	2025 ... 2025 NeurIPS Behavior Challenge 1st place solution
STORM: Search-Guided Generative World Models for Robotic Manipulation	2025-12-20	Show We present STORM (Search-Guided Generative World Models), a novel framework for spatio-temporal reasoning in robotic manipulation that unifies diffusion-based action generation, conditional video prediction, and search-based planning. Unlike prior Vision-Language-Action (VLA) models that rely on abstract latent dynamics or delegate reasoning to language components, STORM grounds planning in explicit visual rollouts, enabling interpretable and foresight-driven decision-making. A diffusion-based VLA policy proposes diverse candidate actions, a generative video world model simulates their visual and reward outcomes, and Monte Carlo Tree Search (MCTS) selectively refines plans through lookahead evaluation. Experiments on the SimplerEnv manipulation benchmark demonstrate that STORM achieves a new state-of-the-art average success rate of 51.0 percent, outperforming strong baselines such as CogACT. Reward-augmented video prediction substantially improves spatio-temporal fidelity and task relevance, reducing Frechet Video Distance by over 75 percent. Moreover, STORM exhibits robust re-planning and failure recovery behavior, highlighting the advantages of search-guided generative world models for long-horizon robotic manipulation.	Under submission
AOMGen: Photoreal, Physics-Consistent Demonstration Generation for Articulated Object Manipulation	2025-12-20	Show Recent advances in Vision-Language-Action (VLA) and world-model methods have improved generalization in tasks such as robotic manipulation and object interaction. However, Successful execution of such tasks depends on large, costly collections of real demonstrations, especially for fine-grained manipulation of articulated objects. To address this, we present AOMGen, a scalable data generation framework for articulated manipulation which is instantiated from a single real scan, demonstration and a library of readily available digital assets, yielding photoreal training data with verified physical states. The framework synthesizes synchronized multi-view RGB temporally aligned with action commands and state annotations for joints and contacts, and systematically varies camera viewpoints, object styles, and object poses to expand a single execution into a diverse corpus. Experimental results demonstrate that fine-tuning VLA policies on AOMGen data increases the success rate from 0% to 88.7%, and the policies are tested on unseen objects and layouts.
Human Centric General Physical Intelligence for Agile Manufacturing Automation	2025-12-20	Show Agile human-centric manufacturing increasingly requires resilient robotic solutions that are capable of safe and productive interactions within unstructured environments of modern factories. While multi-modal sensor fusion provides comprehensive situational awareness yet robots must also contextualize their reasoning to achieve deep semantic understanding of complex scenes. Foundation model particularly Vision-Language-Action (VLA) models have emerged as promising approach on integrating diverse perceptual modalities and spatio-temporal reasoning abilities to ground physical actions to realize General Physical Intelligence (GPI) across various robotic embodiments. Although GPI has been conceptually discussed in literature but its pivotal role and practical deployment in agile manufacturing remain underexplored. To address this gap, this practical review systematically surveys recent advances in VLA models through the lens of GPI by offering comparative analysis of leading implementations and evaluating their industrial readiness via structured ablation study. The state of the art is organized into six thematic pillars including multisensory representation learning, sim2real transfer, planning and control, uncertainty and safety measures and benchmarking. Finally, the review highlights open challenges and future directions for integrating GPI into industrial ecosystems to align with the vision of Industry 5.0 for intelligent, adaptive and collaborative manufacturing ecosystem.	Advan... Advanced Engineering Informatics
cVLA: Towards Efficient Camera-Space VLAs	2025-12-20	Show Vision-Language-Action (VLA) models offer a compelling framework for tackling complex robotic manipulation tasks, but they are often expensive to train. In this paper, we propose a novel VLA approach that leverages the competitive performance of Vision Language Models (VLMs) on 2D images to directly infer robot end-effector poses in image frame coordinates. Unlike prior VLA models that output low-level controls, our model predicts trajectory waypoints, making it both more efficient to train and robot embodiment agnostic. Despite its lightweight design, our next-token prediction architecture effectively learns meaningful and executable robot trajectories. We further explore the underutilized potential of incorporating depth images, inference-time techniques such as decoding strategies, and demonstration-conditioned action generation. Our model is trained on a simulated dataset and exhibits strong sim-to-real transfer capabilities. We evaluate our approach using a combination of simulated and real data, demonstrating its effectiveness on a real robotic system.	20 pa... 20 pages, 10 figures; CoRL 2025 Workshop on Generalizable Priors for Robot Manipulation
Robotic VLA Benefits from Joint Learning with Motion Image Diffusion	2025-12-19	Show Vision-Language-Action (VLA) models have achieved remarkable progress in robotic manipulation by mapping multimodal observations and instructions directly to actions. However, they typically mimic expert trajectories without predictive motion reasoning, which limits their ability to reason about what actions to take. To address this limitation, we propose joint learning with motion image diffusion, a novel strategy that enhances VLA models with motion reasoning capabilities. Our method extends the VLA architecture with a dual-head design: while the action head predicts action chunks as in vanilla VLAs, an additional motion head, implemented as a Diffusion Transformer (DiT), predicts optical-flow-based motion images that capture future dynamics. The two heads are trained jointly, enabling the shared VLM backbone to learn representations that couple robot control with motion knowledge. This joint learning builds temporally coherent and physically grounded representations without modifying the inference pathway of standard VLAs, thereby maintaining test-time latency. Experiments in both simulation and real-world environments demonstrate that joint learning with motion image diffusion improves the success rate of pi-series VLAs to 97.5% on the LIBERO benchmark and 58.0% on the RoboTwin benchmark, yielding a 23% improvement in real-world performance and validating its effectiveness in enhancing the motion reasoning capability of large-scale VLAs.	Websi... Website: https://vla-motion.github.io/
mimic-video: Video-Action Models for Generalizable Robot Control Beyond VLAs	2025-12-19	Show Prevailing Vision-Language-Action Models (VLAs) for robotic manipulation are built upon vision-language backbones pretrained on large-scale, but disconnected static web data. As a result, despite improved semantic generalization, the policy must implicitly infer complex physical dynamics and temporal dependencies solely from robot trajectories. This reliance creates an unsustainable data burden, necessitating continuous, large-scale expert data collection to compensate for the lack of innate physical understanding. We contend that while vision-language pretraining effectively captures semantic priors, it remains blind to physical causality. A more effective paradigm leverages video to jointly capture semantics and visual dynamics during pretraining, thereby isolating the remaining task of low-level control. To this end, we introduce mimic-video, a novel Video-Action Model (VAM) that pairs a pretrained Internet-scale video model with a flow matching-based action decoder conditioned on its latent representations. The decoder serves as an Inverse Dynamics Model (IDM), generating low-level robot actions from the latent representation of video-space action plans. Our extensive evaluation shows that our approach achieves state-of-the-art performance on simulated and real-world robotic manipulation tasks, improving sample efficiency by 10x and convergence speed by 2x compared to traditional VLA architectures.	Revis... Revised Introduction, Related Work, and Appendix. Additional minor notational and grammatical fixes
VLA-AN: An Efficient and Onboard Vision-Language-Action Framework for Aerial Navigation in Complex Environments	2025-12-19	Show This paper proposes VLA-AN, an efficient and onboard Vision-Language-Action (VLA) framework dedicated to autonomous drone navigation in complex environments. VLA-AN addresses four major limitations of existing large aerial navigation models: the data domain gap, insufficient temporal navigation with reasoning, safety issues with generative action policies, and onboard deployment constraints. First, we construct a high-fidelity dataset utilizing 3D Gaussian Splatting (3D-GS) to effectively bridge the domain gap. Second, we introduce a progressive three-stage training framework that sequentially reinforces scene comprehension, core flight skills, and complex navigation capabilities. Third, we design a lightweight, real-time action module coupled with geometric safety correction. This module ensures fast, collision-free, and stable command generation, mitigating the safety risks inherent in stochastic generative policies. Finally, through deep optimization of the onboard deployment pipeline, VLA-AN achieves a robust real-time 8.3x improvement in inference throughput on resource-constrained UAVs. Extensive experiments demonstrate that VLA-AN significantly improves spatial grounding, scene reasoning, and long-horizon navigation, achieving a maximum single-task success rate of 98.1%, and providing an efficient, practical solution for realizing full-chain closed-loop autonomy in lightweight aerial robots.
An Anatomy of Vision-Language-Action Models: From Modules to Milestones and Challenges	2025-12-19	Show Vision-Language-Action (VLA) models are driving a revolution in robotics, enabling machines to understand instructions and interact with the physical world. This field is exploding with new models and datasets, making it both exciting and challenging to keep pace with. This survey offers a clear and structured guide to the VLA landscape. We design it to follow the natural learning path of a researcher: we start with the basic Modules of any VLA model, trace the history through key Milestones, and then dive deep into the core Challenges that define recent research frontier. Our main contribution is a detailed breakdown of the five biggest challenges in: (1) Representation, (2) Execution, (3) Generalization, (4) Safety, and (5) Dataset and Evaluation. This structure mirrors the developmental roadmap of a generalist agent: establishing the fundamental perception-action loop, scaling capabilities across diverse embodiments and environments, and finally ensuring trustworthy deployment-all supported by the essential data infrastructure. For each of them, we review existing approaches and highlight future opportunities. We position this paper as both a foundational guide for newcomers and a strategic roadmap for experienced researchers, with the dual aim of accelerating learning and inspiring new ideas in embodied intelligence. A live version of this survey, with continuous updates, is maintained on our \href{https://suyuz1.github.io/VLA-Survey-Anatomy/}{project page}.	proje... project page: https://suyuz1.github.io/VLA-Survey-Anatomy/
MiVLA: Towards Generalizable Vision-Language-Action Model with Human-Robot Mutual Imitation Pre-training	2025-12-19	Show While leveraging abundant human videos and simulated robot data poses a scalable solution to the scarcity of real-world robot data, the generalization capability of existing vision-language-action models (VLAs) remains limited by mismatches in camera views, visual appearance, and embodiment morphologies. To overcome this limitation, we propose MiVLA, a generalizable VLA empowered by human-robot mutual imitation pre-training, which leverages inherent behavioral similarity between human hands and robotic arms to build a foundation of strong behavioral priors for both human actions and robotic control. Specifically, our method utilizes kinematic rules with left/right hand coordinate systems for bidirectional alignment between human and robot action spaces. Given human or simulated robot demonstrations, MiVLA is trained to forecast behavior trajectories for one embodiment, and imitate behaviors for another one unseen in the demonstration. Based on this mutual imitation, it integrates the behavioral fidelity of real-world human data with the manipulative diversity of simulated robot data into a unified model, thereby enhancing the generalization capability for downstream tasks. Extensive experiments conducted on both simulation and real-world platforms with three robots (ARX, PiPer and LocoMan), demonstrate that MiVLA achieves strong improved generalization capability, outperforming state-of-the-art VLAs (e.g., $\boldsymbolπ_{0}$, $\boldsymbolπ_{0.5}$ and H-RDT) by 25% in simulation, and 14% in real-world robot control tasks.
Phantom Menace: Exploring and Enhancing the Robustness of VLA Models Against Physical Sensor Attacks	2025-12-19	Show Vision-Language-Action (VLA) models revolutionize robotic systems by enabling end-to-end perception-to-action pipelines that integrate multiple sensory modalities, such as visual signals processed by cameras and auditory signals captured by microphones. This multi-modality integration allows VLA models to interpret complex, real-world environments using diverse sensor data streams. Given the fact that VLA-based systems heavily rely on the sensory input, the security of VLA models against physical-world sensor attacks remains critically underexplored. To address this gap, we present the first systematic study of physical sensor attacks against VLAs, quantifying the influence of sensor attacks and investigating the defenses for VLA models. We introduce a novel "Real-Sim-Real" framework that automatically simulates physics-based sensor attack vectors, including six attacks targeting cameras and two targeting microphones, and validates them on real robotic systems. Through large-scale evaluations across various VLA architectures and tasks under varying attack parameters, we demonstrate significant vulnerabilities, with susceptibility patterns that reveal critical dependencies on task types and model designs. We further develop an adversarial-training-based defense that enhances VLA robustness against out-of-distribution physical perturbations caused by sensor attacks while preserving model performance. Our findings expose an urgent need for standardized robustness benchmarks and mitigation strategies to secure VLA deployments in safety-critical environments.	Accep... Accepted by AAAI 2026 main track
GeoPredict: Leveraging Predictive Kinematics and 3D Gaussian Geometry for Precise VLA Manipulation	2025-12-18	Show Vision-Language-Action (VLA) models achieve strong generalization in robotic manipulation but remain largely reactive and 2D-centric, making them unreliable in tasks that require precise 3D reasoning. We propose GeoPredict, a geometry-aware VLA framework that augments a continuous-action policy with predictive kinematic and geometric priors. GeoPredict introduces a trajectory-level module that encodes motion history and predicts multi-step 3D keypoint trajectories of robot arms, and a predictive 3D Gaussian geometry module that forecasts workspace geometry with track-guided refinement along future keypoint trajectories. These predictive modules serve exclusively as training-time supervision through depth-based rendering, while inference requires only lightweight additional query tokens without invoking any 3D decoding. Experiments on RoboCasa Human-50, LIBERO, and real-world manipulation tasks show that GeoPredict consistently outperforms strong VLA baselines, especially in geometry-intensive and spatially demanding scenarios.

robot

Title	Date	Abstract	Comment
Nonlinear Spectral Modeling and Control of Soft-Robotic Muscles from Data	2026-01-06	Show Artificial muscles are essential for compliant musculoskeletal robotics but complicate control due to nonlinear multiphysics dynamics. Hydraulically amplified electrostatic (HASEL) actuators, a class of soft artificial muscles, offer high performance but exhibit memory effects and hysteresis. Here we present a data-driven reduction and control strategy grounded in spectral submanifold (SSM) theory. In the adiabatic regime, where inputs vary slowly relative to intrinsic transients, trajectories rapidly converge to a low-dimensional slow manifold. We learn an explicit input-to-output map on this manifold from forced-response trajectories alone, avoiding decay experiments that can trigger hysteresis. We deploy the SSM-based model for real-time control of an antagonistic HASEL-clutch joint. This approach yields a substantial reduction in tracking error compared to feedback-only and feedforward-only baselines under identical settings. This record-and-control workflow enables rapid characterization and high-performance control of soft muscles and muscle-driven joints without detailed physics-based modeling.
Indicating Robot Vision Capabilities with Augmented Reality	2026-01-06	Show Research indicates that humans can mistakenly assume that robots and humans have the same field of view, possessing an inaccurate mental model of robots. This misperception may lead to failures during human-robot collaboration tasks where robots might be asked to complete impossible tasks about out-of-view objects. The issue is more severe when robots do not have a chance to scan the scene to update their world model while focusing on assigned tasks. To help align humans' mental models of robots' vision capabilities, we propose four field-of-view indicators in augmented reality and conducted a human-subjects experiment (N=41) to evaluate them in a collaborative assembly task regarding accuracy, confidence, task efficiency, and workload. These indicators span a spectrum of positions: two at robot's eye and head space -- deepening eye socket and adding blocks to two sides of the eyes (i.e., egocentric), and two anchoring in the robot's task space -- adding extended blocks from the sides of eyes to the table and placing blocks directly on the tables (i.e., allocentric). Results showed that, when placed directly in the task space, the allocentric indicator yields the highest accuracy, although with a delay in interpreting the robot's field of view. When placed at the robot's eyes, the egocentric indicator of deeper eye sockets, possible for physical alteration, also increased accuracy. In all indicators, participants' confidence was high while cognitive load remained low. Finally, we contribute six guidelines for practitioners to apply our augmented reality indicators or physical alterations to align humans' mental models with robots' vision capabilities.
A High-Fidelity Digital Twin for Robotic Manipulation Based on 3D Gaussian Splatting	2026-01-06	Show Developing high-fidelity, interactive digital twins is crucial for enabling closed-loop motion planning and reliable real-world robot execution, which are essential to advancing sim-to-real transfer. However, existing approaches often suffer from slow reconstruction, limited visual fidelity, and difficulties in converting photorealistic models into planning-ready collision geometry. We present a practical framework that constructs high-quality digital twins within minutes from sparse RGB inputs. Our system employs 3D Gaussian Splatting (3DGS) for fast, photorealistic reconstruction as a unified scene representation. We enhance 3DGS with visibility-aware semantic fusion for accurate 3D labelling and introduce an efficient, filter-based geometry conversion method to produce collision-ready models seamlessly integrated with a Unity-ROS2-MoveIt physics engine. In experiments with a Franka Emika Panda robot performing pick-and-place tasks, we demonstrate that this enhanced geometric accuracy effectively supports robust manipulation in real-world trials. These results demonstrate that 3DGS-based digital twins, enriched with semantic and geometric consistency, offer a fast, reliable, and scalable path from perception to manipulation in unstructured environments.	Under... Under review of Journal of Robot Learning
HEXAR: a Hierarchical Explainability Architecture for Robots	2026-01-06	Show As robotic systems become increasingly complex, the need for explainable decision-making becomes critical. Existing explainability approaches in robotics typically either focus on individual modules, which can be difficult to query from the perspective of high-level behaviour, or employ monolithic approaches, which do not exploit the modularity of robotic architectures. We present HEXAR (Hierarchical EXplainability Architecture for Robots), a novel framework that provides a plug-in, hierarchical approach to generate explanations about robotic systems. HEXAR consists of specialised component explainers using diverse explanation techniques (e.g., LLM-based reasoning, causal models, feature importance, etc) tailored to specific robot modules, orchestrated by an explainer selector that chooses the most appropriate one for a given query. We implement and evaluate HEXAR on a TIAGo robot performing assistive tasks in a home environment, comparing it against end-to-end and aggregated baseline approaches across 180 scenario-query variations. We observe that HEXAR significantly outperforms baselines in root cause identification, incorrect information exclusion, and runtime, offering a promising direction for transparent autonomous systems.	8 pages, 6 figures
Validating Generalist Robots with Situation Calculus and STL Falsification	2026-01-06	Show Generalist robots are becoming a reality, capable of interpreting natural language instructions and executing diverse operations. However, their validation remains challenging because each task induces its own operational context and correctness specification, exceeding the assumptions of traditional validation methods. We propose a two-layer validation framework that combines abstract reasoning with concrete system falsification. At the abstract layer, situation calculus models the world and derives weakest preconditions, enabling constraint-aware combinatorial testing to systematically generate diverse, semantically valid world-task configurations with controllable coverage strength. At the concrete layer, these configurations are instantiated for simulation-based falsification with STL monitoring. Experiments on tabletop manipulation tasks show that our framework effectively uncovers failure cases in the NVIDIA GR00T controller, demonstrating its promise for validating general-purpose robot autonomy.
Evaluating Gemini Robotics Policies in a Veo World Simulator	2026-01-06	Show Generative world models hold significant potential for simulating interactions with visuomotor policies in varied environments. Frontier video models can enable generation of realistic observations and environment interactions in a scalable and general manner. However, the use of video models in robotics has been limited primarily to in-distribution evaluations, i.e., scenarios that are similar to ones used to train the policy or fine-tune the base video model. In this report, we demonstrate that video models can be used for the entire spectrum of policy evaluation use cases in robotics: from assessing nominal performance to out-of-distribution (OOD) generalization, and probing physical and semantic safety. We introduce a generative evaluation system built upon a frontier video foundation model (Veo). The system is optimized to support robot action conditioning and multi-view consistency, while integrating generative image-editing and multi-view completion to synthesize realistic variations of real-world scenes along multiple axes of generalization. We demonstrate that the system preserves the base capabilities of the video model to enable accurate simulation of scenes that have been edited to include novel interaction objects, novel visual backgrounds, and novel distractor objects. This fidelity enables accurately predicting the relative performance of different policies in both nominal and OOD conditions, determining the relative impact of different axes of generalization on policy performance, and performing red teaming of policies to expose behaviors that violate physical or semantic safety constraints. We validate these capabilities through 1600+ real-world evaluations of eight Gemini Robotics policy checkpoints and five tasks for a bimanual manipulator.
DDBot: Differentiable Physics-based Digging Robot for Unknown Granular Materials	2026-01-06	Show Automating the manipulation of granular materials poses significant challenges due to complex contact dynamics, unpredictable material properties, and intricate system states. Existing approaches often fail to achieve efficiency and accuracy in such tasks. To fill the research gap, this article studies the small-scale and high-precision granular material digging task with unknown physical properties. A key scientific problem addressed is the feasibility of applying first-order gradient-based optimization to complex differentiable granular material simulation and overcoming associated numerical instability. A new framework, named differentiable digging robot (DDBot), is proposed to manipulate granular materials, including sand and soil. Specifically, we equip DDBot with a differentiable physics-based simulator, tailored for granular material manipulation, powered by GPU-accelerated parallel computing and automatic differentiation. DDBot can perform efficient differentiable system identification and high-precision digging skill optimization for unknown granular materials, which is enabled by a differentiable skill-to-action mapping, a task-oriented demonstration method, gradient clipping and line search-based gradient descent. Experimental results show that DDBot can efficiently (converge within 5 to 20 minutes) identify unknown granular material dynamics and optimize digging skills, with high-precision results in zero-shot real-world deployments, highlighting its practicality. Benchmark results against state-of-the-art baselines also confirm the robustness and efficiency of DDBot in such digging tasks.	Publi... Published as a regular paper by the IEEE Transactions on Robotics
Augmented Reality for RObots (ARRO): Pointing Visuomotor Policies Towards Visual Robustness	2026-01-06	Show Visuomotor policies trained on human expert demonstrations have recently shown strong performance across a wide range of robotic manipulation tasks. However, these policies remain highly sensitive to domain shifts stemming from background or robot embodiment changes, which limits their generalization capabilities. In this paper, we present ARRO, a novel visual representation that leverages zero-shot open-vocabulary segmentation and object detection models to efficiently mask out task-irrelevant regions of the scene in real time without requiring additional training, modeling of the setup, or camera calibration. By filtering visual distractors and overlaying virtual guides during both training and inference, ARRO improves robustness to scene variations and reduces the need for additional data collection. We extensively evaluate ARRO with Diffusion Policy on a range of tabletop manipulation tasks in both simulation and real-world environments, and further demonstrate its compatibility and effectiveness with generalist robot policies, such as Octo, OpenVLA and Pi Zero. Across all settings in our evaluation, ARRO yields consistent performance gains, allows for selective masking to choose between different objects, and shows robustness even to challenging segmentation conditions. Videos showcasing our results are available at: https://augmented-reality-for-robots.github.io/
Steering Flexible Linear Objects in Planar Environments by Two Robot Hands Using Euler's Elastica Solutions	2026-01-06	Show The manipulation of flexible objects such as cables, wires and fresh food items by robot hands forms a special challenge in robot grasp mechanics. This paper considers the steering of flexible linear objects in planar environments by two robot hands. The flexible linear object, modeled as an elastic non-stretchable rod, is manipulated by varying the gripping endpoint positions while keeping equal endpoint tangents. The flexible linear object shape has a closed form solution in terms of the grasp endpoint positions and tangents, called Euler's elastica. This paper obtains the elastica solutions under the optimal control framework, then uses the elastica solutions to obtain closed-form criteria for non self-intersection, stability and obstacle avoidance of the flexible linear object. The new tools are incorporated into a planning scheme for steering flexible linear objects in planar environments populated by sparsely spaced obstacles. The scheme is fully implemented and demonstrated with detailed examples.
Reinforcement Learning for Follow-the-Leader Robotic Endoscopic Navigation via Synthetic Data	2026-01-06	Show Autonomous navigation is crucial for both medical and industrial endoscopic robots, enabling safe and efficient exploration of narrow tubular environments without continuous human intervention, where avoiding contact with the inner walls has been a longstanding challenge for prior approaches. We present a follow-the-leader endoscopic robot based on a flexible continuum structure designed to minimize contact between the endoscope body and intestinal walls, thereby reducing patient discomfort. To achieve this objective, we propose a vision-based deep reinforcement learning framework guided by monocular depth estimation. A realistic intestinal simulation environment was constructed in \textit{NVIDIA Omniverse} to train and evaluate autonomous navigation strategies. Furthermore, thousands of synthetic intraluminal images were generated using NVIDIA Replicator to fine-tune the Depth Anything model, enabling dense three-dimensional perception of the intestinal environment with a single monocular camera. Subsequently, we introduce a geometry-aware reward and penalty mechanism to enable accurate lumen tracking. Compared with the original Depth Anything model, our method improves $δ_{1}$ depth accuracy by 39.2% and reduces the navigation J-index by 0.67 relative to the second-best method, demonstrating the robustness and effectiveness of the proposed approach.
Efficient Swept Volume-Based Trajectory Generation for Arbitrary-Shaped Ground Robot Navigation	2026-01-06	Show Navigating an arbitrary-shaped ground robot safely in cluttered environments remains a challenging problem. The existing trajectory planners that account for the robot's physical geometry severely suffer from the intractable runtime. To achieve both computational efficiency and Continuous Collision Avoidance (CCA) of arbitrary-shaped ground robot planning, we proposed a novel coarse-to-fine navigation framework that significantly accelerates planning. In the first stage, a sampling-based method selectively generates distinct topological paths that guarantee a minimum inflated margin. In the second stage, a geometry-aware front-end strategy is designed to discretize these topologies into full-state robot motion sequences while concurrently partitioning the paths into SE(2) sub-problems and simpler R2 sub-problems for back-end optimization. In the final stage, an SVSDF-based optimizer generates trajectories tailored to these sub-problems and seamlessly splices them into a continuous final motion plan. Extensive benchmark comparisons show that the proposed method is one to several orders of magnitude faster than the cutting-edge methods in runtime while maintaining a high planning success rate and ensuring CCA.
Advancing Assistive Robotics: Multi-Modal Navigation and Biophysical Monitoring for Next-Generation Wheelchairs	2026-01-06	Show Assistive electric-powered wheelchairs (EPWs) have become essential mobility aids for people with disabilities such as amyotrophic lateral sclerosis (ALS), post-stroke hemiplegia, and dementia-related mobility impairment. This work presents a novel multi-modal EPW control system designed to prioritize patient needs while allowing seamless switching between control modes. Four complementary interfaces, namely joystick, speech, hand gesture, and electrooculography (EOG), are integrated with a continuous vital sign monitoring framework measuring heart rate variability, oxygen saturation (SpO2), and skin temperature. This combination enables greater patient independence while allowing caregivers to maintain real-time supervision and early intervention capability. Two-point calibration of the biophysical sensors against clinical reference devices resulted in root mean square errors of at most 2 bpm for heart rate, 0.5 degree Celsius for skin temperature, and 1 percent for SpO2. Experimental evaluation involved twenty participants with mobility impairments executing a total of 500 indoor navigation commands. The achieved command recognition accuracies were 99 percent for joystick control, 97 percent plus or minus 2 percent for speech, and 95 percent plus or minus 3 percent for hand gesture, with an average closed-loop latency of 20 plus or minus 0.5 milliseconds. Caregivers receive real-time alerts through an Android application following encrypted cloud transmission of physiological data. By integrating multi-modal mobility control with cloud-enabled health monitoring and reporting latency and energy budgets, the proposed prototype addresses key challenges in assistive robotics, contributes toward compliance with ISO 7176-31 and IEC 80601-2-78 safety standards, and establishes a foundation for future adaptive machine learning enhancements.
RoboTracer: Mastering Spatial Trace with Reasoning in Vision-Language Models for Robotics	2026-01-06	Show Spatial tracing, as a fundamental embodied interaction ability for robots, is inherently challenging as it requires multi-step metric-grounded reasoning compounded with complex spatial referring and real-world metric measurement. However, existing methods struggle with this compositional task. To this end, we propose RoboTracer, a 3D-aware VLM that first achieves both 3D spatial referring and measuring via a universal spatial encoder and a regression-supervised decoder to enhance scale awareness during supervised fine-tuning (SFT). Moreover, RoboTracer advances multi-step metric-grounded reasoning via reinforcement fine-tuning (RFT) with metric-sensitive process rewards, supervising key intermediate perceptual cues to accurately generate spatial traces. To support SFT and RFT training, we introduce TraceSpatial, a large-scale dataset of 30M QA pairs, spanning outdoor/indoor/tabletop scenes and supporting complex reasoning processes (up to 9 steps). We further present TraceSpatial-Bench, a challenging benchmark filling the gap to evaluate spatial tracing. Experimental results show that RoboTracer surpasses baselines in spatial understanding, measuring, and referring, with an average success rate of 79.1%, and also achieves SOTA performance on TraceSpatial-Bench by a large margin, exceeding Gemini-2.5-Pro by 36% accuracy. Notably, RoboTracer can be integrated with various control policies to execute long-horizon, dynamic tasks across diverse robots (UR5, G1 humanoid) in cluttered real-world scenes. See the project page at https://zhoues.github.io/RoboTracer.	Proje... Project page: https://zhoues.github.io/RoboTracer
Learning to Nudge: A Scalable Barrier Function Framework for Safe Robot Interaction in Dense Clutter	2026-01-06	Show Robots operating in everyday environments must navigate and manipulate within densely cluttered spaces, where physical contact with surrounding objects is unavoidable. Traditional safety frameworks treat contact as unsafe, restricting robots to collision avoidance and limiting their ability to function in dense, everyday settings. As the number of objects grows, model-based approaches for safe manipulation become computationally intractable; meanwhile, learned methods typically tie safety to the task at hand, making them hard to transfer to new tasks without retraining. In this work we introduce Dense Contact Barrier Functions(DCBF). Our approach bypasses the computational complexity of explicitly modeling multi-object dynamics by instead learning a composable, object-centric function that implicitly captures the safety constraints arising from physical interactions. Trained offline on interactions with a few objects, the learned DCBFcomposes across arbitrary object sets at runtime, producing a single global safety filter that scales linearly and transfers across tasks without retraining. We validate our approach through simulated experiments in dense clutter, demonstrating its ability to enable collision-free navigation and safe, contact-rich interaction in suitable settings.
InternVLA-A1: Unifying Understanding, Generation and Action for Robotic Manipulation	2026-01-05	Show Prevalent Vision-Language-Action (VLA) models are typically built upon Multimodal Large Language Models (MLLMs) and demonstrate exceptional proficiency in semantic understanding, but they inherently lack the capability to deduce physical world dynamics. Consequently, recent approaches have shifted toward World Models, typically formulated via video prediction; however, these methods often suffer from a lack of semantic grounding and exhibit brittleness when handling prediction errors. To synergize semantic understanding with dynamic predictive capabilities, we present InternVLA-A1. This model employs a unified Mixture-of-Transformers architecture, coordinating three experts for scene understanding, visual foresight generation, and action execution. These components interact seamlessly through a unified masked self-attention mechanism. Building upon InternVL3 and Qwen3-VL, we instantiate InternVLA-A1 at 2B and 3B parameter scales. We pre-train these models on hybrid synthetic-real datasets spanning InternData-A1 and Agibot-World, covering over 533M frames. This hybrid training strategy effectively harnesses the diversity of synthetic simulation data while minimizing the sim-to-real gap. We evaluated InternVLA-A1 across 12 real-world robotic tasks and simulation benchmark. It significantly outperforms leading models like pi0 and GR00T N1.5, achieving a 14.5% improvement in daily tasks and a 40%-73.3% boost in dynamic settings, such as conveyor belt sorting.	Homep... Homepage: https://internrobotics.github.io/internvla-a1.github.io/
From Chat Control to Robot Control: The Backdoors Left Open for the Sake of Safety	2026-01-05	Show This paper explores how a recent European Union proposal, the so-called Chat Control law, which creates regulatory incentives for providers to implement content detection and communication scanning, could transform the foundations of human-robot interaction (HRI). As robots increasingly act as interpersonal communication channels in care, education, and telepresence, they convey not only speech but also gesture, emotion, and contextual cues. We argue that extending digital surveillance laws to such embodied systems would entail continuous monitoring, embedding observation into the very design of everyday robots. This regulation blurs the line between protection and control, turning companions into potential informants. At the same time, monitoring mechanisms that undermine end-to-end encryption function as de facto backdoors, expanding the attack surface and allowing adversaries to exploit legally induced monitoring infrastructures. This creates a paradox of safety through insecurity: systems introduced to protect users may instead compromise their privacy, autonomy, and trust. This work does not aim to predict the future, but to raise awareness and help prevent certain futures from materialising.	15 pages, 2 figures
SingingBot: An Avatar-Driven System for Robotic Face Singing Performance	2026-01-05	Show Equipping robotic faces with singing capabilities is crucial for empathetic Human-Robot Interaction. However, existing robotic face driving research primarily focuses on conversations or mimicking static expressions, struggling to meet the high demands for continuous emotional expression and coherence in singing. To address this, we propose a novel avatar-driven framework for appealing robotic singing. We first leverage portrait video generation models embedded with extensive human priors to synthesize vivid singing avatars, providing reliable expression and emotion guidance. Subsequently, these facial features are transferred to the robot via semantic-oriented mapping functions that span a wide expression space. Furthermore, to quantitatively evaluate the emotional richness of robotic singing, we propose the Emotion Dynamic Range metric to measure the emotional breadth within the Valence-Arousal space, revealing that a broad emotional spectrum is crucial for appealing performances. Comprehensive experiments prove that our method achieves rich emotional expressions while maintaining lip-audio synchronization, significantly outperforming existing approaches.
Vision-Based Early Fault Diagnosis and Self-Recovery for Strawberry Harvesting Robots	2026-01-05	Show Strawberry harvesting robots faced persistent challenges such as low integration of visual perception, fruit-gripper misalignment, empty grasping, and strawberry slippage from the gripper due to insufficient gripping force, all of which compromised harvesting stability and efficiency in orchard environments. To overcome these issues, this paper proposed a visual fault diagnosis and self-recovery framework that integrated multi-task perception with corrective control strategies. At the core of this framework was SRR-Net, an end-to-end multi-task perception model that simultaneously performed strawberry detection, segmentation, and ripeness estimation, thereby unifying visual perception with fault diagnosis. Based on this integrated perception, a relative error compensation method based on the simultaneous target-gripper detection was designed to address positional misalignment, correcting deviations when error exceeded the tolerance threshold. To mitigate empty grasping and fruit-slippage faults, an early abort strategy was implemented. A micro-optical camera embedded in the end-effector provided real-time visual feedback, enabling grasp detection during the deflating stage and strawberry slip prediction during snap-off through MobileNet V3-Small classifier and a time-series LSTM classifier. Experiments demonstrated that SRR-Net maintained high perception accuracy. For detection, it achieved a precision of 0.895 and recall of 0.813 on strawberries, and 0.972/0.958 on hands. In segmentation, it yielded a precision of 0.887 and recall of 0.747 for strawberries, and 0.974/0.947 for hands. For ripeness estimation, SRR-Net attained a mean absolute error of 0.035, while simultaneously supporting multi-task perception and sustaining a competitive inference speed of 163.35 FPS.
Genie Sim 3.0 : A High-Fidelity Comprehensive Simulation Platform for Humanoid Robot	2026-01-05	Show The development of robust and generalizable robot learning models is critically contingent upon the availability of large-scale, diverse training data and reliable evaluation benchmarks. Collecting data in the physical world poses prohibitive costs and scalability challenges, and prevailing simulation benchmarks frequently suffer from fragmentation, narrow scope, or insufficient fidelity to enable effective sim-to-real transfer. To address these challenges, we introduce Genie Sim 3.0, a unified simulation platform for robotic manipulation. We present Genie Sim Generator, a large language model (LLM)-powered tool that constructs high-fidelity scenes from natural language instructions. Its principal strength resides in rapid and multi-dimensional generalization, facilitating the synthesis of diverse environments to support scalable data collection and robust policy evaluation. We introduce the first benchmark that pioneers the application of LLM for automated evaluation. It leverages LLM to mass-generate evaluation scenarios and employs Vision-Language Model (VLM) to establish an automated assessment pipeline. We also release an open-source dataset comprising more than 10,000 hours of synthetic data across over 200 tasks. Through systematic experimentation, we validate the robust zero-shot sim-to-real transfer capability of our open-source dataset, demonstrating that synthetic data can server as an effective substitute for real-world data under controlled conditions for scalable policy training. For code and dataset details, please refer to: https://github.com/AgibotTech/genie_sim.
Learning Diffusion Policy from Primitive Skills for Robot Manipulation	2026-01-05	Show Diffusion policies (DP) have recently shown great promise for generating actions in robotic manipulation. However, existing approaches often rely on global instructions to produce short-term control signals, which can result in misalignment in action generation. We conjecture that the primitive skills, referred to as fine-grained, short-horizon manipulations, such as move up'' and open the gripper'', provide a more intuitive and effective interface for robot learning. To bridge this gap, we propose SDP, a skill-conditioned DP that integrates interpretable skill learning with conditional action planning. SDP abstracts eight reusable primitive skills across tasks and employs a vision-language model to extract discrete representations from visual observations and language instructions. Based on them, a lightweight router network is designed to assign a desired primitive skill for each state, which helps construct a single-skill policy to generate skill-aligned actions. By decomposing complex tasks into a sequence of primitive skills and selecting a single-skill policy, SDP ensures skill-consistent behavior across diverse tasks. Extensive experiments on two challenging simulation benchmarks and real-world robot deployments demonstrate that SDP consistently outperforms SOTA methods, providing a new paradigm for skill-based robot learning with diffusion policies.	Accepted to AAAI2026
From Metrics to Meaning: Insights from a Mixed-Methods Field Experiment on Retail Robot Deployment	2026-01-05	Show We report a mixed-methods field experiment of a conversational service robot deployed under everyday staffing discretion in a live bedding store. Over 12 days we alternated three conditions--Baseline (no robot), Robot-only, and Robot+Fixture--and video-annotated the service funnel from passersby to purchase. An explanatory sequential design then used six post-experiment staff interviews to interpret the quantitative patterns. Quantitatively, the robot increased stopping per passerby (highest with the fixture), yet clerk-led downstream steps per stopper--clerk approach, store entry, assisted experience, and purchase--decreased. Interviews explained this divergence: clerks avoided interrupting ongoing robot-customer talk, struggled with ambiguous timing amid conversational latency, and noted child-centered attraction that often satisfied curiosity at the doorway. The fixture amplified visibility but also anchored encounters at the threshold, creating a well-defined micro-space where needs could ``close'' without moving inside. We synthesize these strands into an integrative account from the initial show of interest on the part of a customer to their entering the store and derive actionable guidance. The results advance the understanding of interactions between customers, staff members, and the robot and offer practical recommendations for deploying service robots in high-touch retail.
Multi-Robot Data-Free Continual Communicative Learning (CCL) from Black-Box Visual Place Recognition Models	2026-01-05	Show In emerging multi-robot societies, heterogeneous agents must continually extract and integrate local knowledge from one another through communication, even when their internal models are completely opaque. Existing approaches to continual or collaborative learning for visual place recognition (VPR) largely assume white-box access to model parameters or shared training datasets, which is unrealistic when robots encounter unknown peers in the wild. This paper introduces \emph{Continual Communicative Learning (CCL)}, a data-free multi-robot framework in which a traveler robot (student) continually improves its VPR capability by communicating with black-box teacher models via a constrained query--response channel. We repurpose Membership Inference Attacks (MIA), originally developed as privacy attacks on machine learning models, as a constructive communication primitive to reconstruct pseudo-training sets from black-box VPR teachers without accessing their parameters or raw data. To overcome the intrinsic communication bottleneck caused by the low sampling efficiency of black-box MIA, we propose a prior-based query strategy that leverages the student's own VPR prior to focus queries on informative regions of the embedding space, thereby reducing the knowledge transfer (KT) cost. Experimental results on a standard multi-session VPR benchmark demonstrate that the proposed CCL framework yields substantial performance gains for low-performing robots under modest communication budgets, highlighting CCL as a promising building block for scalable and fault-tolerant multi-robot systems.	6 pag... 6 pages, 4 figures, technical report
RoboBPP: Benchmarking Robotic Online Bin Packing with Physics-based Simulation	2026-01-05	Show Physical feasibility in 3D bin packing is a key requirement in modern industrial logistics and robotic automation. With the growing adoption of industrial automation, online bin packing has gained increasing attention. However, inconsistencies in problem settings, test datasets, and evaluation metrics have hindered progress in the field, and there is a lack of a comprehensive benchmarking system. Direct testing on real hardware is costly, and building a realistic simulation environment is also challenging. To address these limitations, we introduce RoboBPP, a benchmarking system designed for robotic online bin packing. RoboBPP integrates a physics-based simulator to assess physical feasibility. In our simulation environment, we introduce a robotic arm and boxes at real-world scales to replicate real industrial packing workflows. By simulating conditions that arise in real industrial applications, we ensure that evaluated algorithms are practically deployable. In addition, prior studies often rely on synthetic datasets whose distributions differ from real-world industrial data. To address this issue, we collect three datasets from real industrial workflows, including assembly-line production, logistics packing, and furniture manufacturing. The benchmark comprises three carefully designed test settings and extends existing evaluation metrics with new metrics for structural stability and operational safety. We design a scoring system and derive a range of insights from the evaluation results. RoboBPP is fully open-source and is equipped with visualization tools and an online leaderboard, providing a reproducible and extensible foundation for future research and industrial applications (https://robot-bin-packing-benchmark.github.io).	Under... Under review at the International Journal of Robotics Research (IJRR)
CausalNav: A Long-term Embodied Navigation System for Autonomous Mobile Robots in Dynamic Outdoor Scenarios	2026-01-05	Show Autonomous language-guided navigation in large-scale outdoor environments remains a key challenge in mobile robotics, due to difficulties in semantic reasoning, dynamic conditions, and long-term stability. We propose CausalNav, the first scene graph-based semantic navigation framework tailored for dynamic outdoor environments. We construct a multi-level semantic scene graph using LLMs, referred to as the Embodied Graph, that hierarchically integrates coarse-grained map data with fine-grained object entities. The constructed graph serves as a retrievable knowledge base for Retrieval-Augmented Generation (RAG), enabling semantic navigation and long-range planning under open-vocabulary queries. By fusing real-time perception with offline map data, the Embodied Graph supports robust navigation across varying spatial granularities in dynamic outdoor environments. Dynamic objects are explicitly handled in both the scene graph construction and hierarchical planning modules. The Embodied Graph is continuously updated within a temporal window to reflect environmental changes and support real-time semantic navigation. Extensive experiments in both simulation and real-world settings demonstrate superior robustness and efficiency.	Accep... Accepted by IEEE Robotics and Automation Letters (RA-L)
RNBF: Real-Time RGB-D Based Neural Barrier Functions for Safe Robotic Navigation	2026-01-05	Show Autonomous safe navigation in unstructured and novel environments poses significant challenges, especially when environment information can only be provided through low-cost vision sensors. Although safe reactive approaches have been proposed to ensure robot safety in complex environments, many base their theory off the assumption that the robot has prior knowledge on obstacle locations and geometries. In this paper, we present a real-time, vision-based framework that constructs continuous, first-order differentiable Signed Distance Fields (SDFs) of unknown environments entirely online, without any pre-training, and is fully compatible with established SDF-based reactive controllers. To achieve robust performance under practical sensing conditions, our approach explicitly accounts for noise in affordable RGB-D cameras, refining the neural SDF representation online for smoother geometry and stable gradient estimates. We validate the proposed method in simulation and real-world experiments using a Fetch robot. Videos and supplementary material are available at https://satyajeetburla.github.io/rnbf/.
SurgWorld: Learning Surgical Robot Policies from Videos via World Modeling	2026-01-05	Show Data scarcity remains a fundamental barrier to achieving fully autonomous surgical robots. While large scale vision language action (VLA) models have shown impressive generalization in household and industrial manipulation by leveraging paired video action data from diverse domains, surgical robotics suffers from the paucity of datasets that include both visual observations and accurate robot kinematics. In contrast, vast corpora of surgical videos exist, but they lack corresponding action labels, preventing direct application of imitation learning or VLA training. In this work, we aim to alleviate this problem by learning policy models from SurgWorld, a world model designed for surgical physical AI. We curated the Surgical Action Text Alignment (SATA) dataset with detailed action description specifically for surgical robots. Then we built SurgeWorld based on the most advanced physical AI world model and SATA. It's able to generate diverse, generalizable and realistic surgery videos. We are also the first to use an inverse dynamics model to infer pseudokinematics from synthetic surgical videos, producing synthetic paired video action data. We demonstrate that a surgical VLA policy trained with these augmented data significantly outperforms models trained only on real demonstrations on a real surgical robot platform. Our approach offers a scalable path toward autonomous surgical skill acquisition by leveraging the abundance of unlabeled surgical video and generative world modeling, thus opening the door to generalizable and data efficient surgical robot policies.
Explicit World Models for Reliable Human-Robot Collaboration	2026-01-05	Show This paper addresses the topic of robustness under sensing noise, ambiguous instructions, and human-robot interaction. We take a radically different tack to the issue of reliable embodied AI: instead of focusing on formal verification methods aimed at achieving model predictability and robustness, we emphasise the dynamic, ambiguous and subjective nature of human-robot interactions that requires embodied AI systems to perceive, interpret, and respond to human intentions in a manner that is consistent, comprehensible and aligned with human expectations. We argue that when embodied agents operate in human environments that are inherently social, multimodal, and fluid, reliability is contextually determined and only has meaning in relation to the goals and expectations of humans involved in the interaction. This calls for a fundamentally different approach to achieving reliable embodied AI that is centred on building and updating an accessible "explicit world model" representing the common ground between human and AI, that is used to align robot behaviours with human expectations.	Accep... Accepted to AAAI-26 Bridge Program B10: Making Embodied AI Reliable with Testing and Formal Verification
How Robot Dogs See the Unseeable: Improving Visual Interpretability via Peering for Exploratory Robots	2026-01-04	Show In vegetated environments, such as forests, exploratory robots play a vital role in navigating complex, cluttered environments where human access is limited and traditional equipment struggles. Visual occlusion from obstacles, such as foliage, can severely obstruct a robot's sensors, impairing scene understanding. We show that "peering", a characteristic side-to-side movement used by insects to overcome their visual limitations, can also allow robots to markedly improve visual reasoning under partial occlusion. This is accomplished by applying core signal processing principles, specifically optical synthetic aperture sensing, together with the vision reasoning capabilities of modern large multimodal models. Peering enables real-time, high-resolution, and wavelength-independent perception, which is crucial for vision-based scene understanding across a wide range of applications. The approach is low-cost and immediately deployable on any camera-equipped robot. We investigated different peering motions and occlusion masking strategies, demonstrating that, unlike peering, state-of-the-art multi-view 3D vision techniques fail in these conditions due to their high susceptibility to occlusion. Our experiments were carried out on an industrial-grade quadrupedal robot. However, the ability to peer is not limited to such platforms, but potentially also applicable to bipedal, hexapod, wheeled, or crawling platforms. Robots that can effectively see through partial occlusion will gain superior perception abilities - including enhanced scene understanding, situational awareness, camouflage breaking, and advanced navigation in complex environments.
Action-Sketcher: From Reasoning to Action via Visual Sketches for Long-Horizon Robotic Manipulation	2026-01-04	Show Long-horizon robotic manipulation is increasingly important for real-world deployment, requiring spatial disambiguation in complex layouts and temporal resilience under dynamic interaction. However, existing end-to-end and hierarchical Vision-Language-Action (VLA) policies often rely on text-only cues while keeping plan intent latent, which undermines referential grounding in cluttered or underspecified scenes, impedes effective task decomposition of long-horizon goals with close-loop interaction, and limits causal explanation by obscuring the rationale behind action choices. To address these issues, we first introduce Visual Sketch, an implausible visual intermediate that renders points, boxes, arrows, and typed relations in the robot's current views to externalize spatial intent, connect language to scene geometry. Building on Visual Sketch, we present Action-Sketcher, a VLA framework that operates in a cyclic See-Think-Sketch-Act workflow coordinated by adaptive token-gated strategy for reasoning triggers, sketch revision, and action issuance, thereby supporting reactive corrections and human interaction while preserving real-time action prediction. To enable scalable training and evaluation, we curate diverse corpus with interleaved images, text, Visual Sketch supervision, and action sequences, and train Action-Sketcher with a multi-stage curriculum recipe that combines interleaved sequence alignment for modality unification, language-to-sketch consistency for precise linguistic grounding, and imitation learning augmented with sketch-to-action reinforcement for robustness. Extensive experiments on cluttered scenes and multi-object tasks, in simulation and on real-world tasks, show improved long-horizon success, stronger robustness to dynamic scene changes, and enhanced interpretability via editable sketches and step-wise plans. Project website: https://action-sketcher.github.io	26 pages, 14 figures
Standing Tall: Sim to Real Fall Classification and Lead Time Prediction for Bipedal Robots	2026-01-04	Show This paper extends a previously proposed fall prediction algorithm to a real-time (online) setting, with implementations in both hardware and simulation. The system is validated on the full-sized bipedal robot Digit, where the real-time version achieves performance comparable to the offline implementation while maintaining a zero false positive rate, an average lead time (defined as the difference between the true and predicted fall time) of 1.1s (well above the required minimum of 0.2s), and a maximum lead time error of just 0.03s. It also achieves a high recovery rate of 0.97, demonstrating its effectiveness in real-world deployment. In addition to the real-time implementation, this work identifies key limitations of the original algorithm, particularly under omnidirectional faults, and introduces a fine-tuned strategy to improve robustness. The enhanced algorithm shows measurable improvements across all evaluated metrics, including a 0.05 reduction in average false positive rate and a 1.19s decrease in the maximum error of the average predicted lead time.	This ... This work has been submitted to the IEEE for possible publication
AIMS: An Adaptive Integration of Multi-Sensor Measurements for Quadrupedal Robot Localization	2026-01-04	Show This paper addresses the problem of accurate localization for quadrupedal robots operating in narrow tunnel-like environments. Due to the long and homogeneous characteristics of such scenarios, LiDAR measurements often provide weak geometric constraints, making traditional sensor fusion methods susceptible to accumulated motion estimation errors. To address these challenges, we propose AIMS, an adaptive LiDAR-IMU-leg odometry fusion method for robust quadrupedal robot localization in degenerate environments. The proposed method is formulated within an error-state Kalman filtering framework, where LiDAR and leg odometry measurements are integrated with IMU-based state prediction, and measurement noise covariance matrices are adaptively adjusted based on online degeneracy-aware reliability assessment. Experimental results obtained in narrow corridor environments demonstrate that the proposed method improves localization accuracy and robustness compared with state-of-the-art approaches.
H2R: A Human-to-Robot Data Augmentation for Robot Pre-training from Videos	2026-01-04	Show Large-scale pre-training using videos has proven effective for robot learning. However, the models pre-trained on such data can be suboptimal for robot learning due to the significant visual gap between human hands and those of different robots. To remedy this, we propose H2R, a simple data augmentation technique that detects human hand keypoints, synthesizes robot motions in simulation, and composites rendered robots into egocentric videos. This process explicitly bridges the visual gap between human and robot embodiments during pre-training. We apply H2R to augment large-scale egocentric human video datasets such as Ego4D and SSv2, replacing human hands with simulated robotic arms to generate robot-centric training data. Based on this, we construct and release a family of 1M-scale datasets covering multiple robot embodiments (UR5 with gripper/Leaphand, Franka) and data sources (SSv2, Ego4D). To verify the effectiveness of the augmentation pipeline, we introduce a CLIP-based image-text similarity metric that quantitatively evaluates the semantic fidelity of robot-rendered frames to the original human actions. We validate H2R across three simulation benchmarks: Robomimic, RLBench and PushT and real-world manipulation tasks with a UR5 robot equipped with Gripper and Leaphand end-effectors. H2R consistently improves downstream success rates, yielding gains of 5.0%-10.2% in simulation and 6.7%-23.3% in real-world tasks across various visual encoders and policy learning methods. These results indicate that H2R improves the generalization ability of robotic policies by mitigating the visual discrepancies between human and robot domains.
Sampling Strategy Design for Model Predictive Path Integral Control on Legged Robot Locomotion	2026-01-04	Show Model Predictive Path Integral (MPPI) control has emerged as a powerful sampling-based optimal control method for complex, nonlinear, and high-dimensional systems. However, directly applying MPPI to legged robotic systems presents several challenges. This paper systematically investigates the role of sampling strategy design within the MPPI framework for legged robot locomotion. Based upon the idea of structured control parameterization, we explore and compare multiple sampling strategies within the framework, including both unstructured and spline-based approaches. Through extensive simulations on a quadruped robot platform, we evaluate how different sampling strategies affect control smoothness, task performance, robustness, and sample efficiency. The results provide new insights into the practical implications of sampling design for deploying MPPI on complex legged systems.
Bridging Language Gaps: Utilizing Interactive Robots to Teach Cantonese in Real-Life Contexts for Newly-Arrived Children	2026-01-03	Show Hong Kong's education system is notably multicultural, including local, non-Chinese-speaking, and newly arrived students (NAS) (Mandarine Chinese-speaking). NAS can guess the meaning of vocabulary but cannot speak out, presenting unique challenges for them, particularly language barriers and cultural differences. These challenges hinder their academic success and social integration, leading to feelings of isolation and demotivation. Current resources often fail to address the emotional well-being of these students and predominantly focus on English language acquisition, leaving a gap in support for learning Cantonese and navigating the local cultural landscape. This study explores the effectiveness of an interactive robot, Boon Boon, in teaching Cantonese through real-life contexts to enhance NAS children learning engagement and motivation. The research questions are: (1) How does interactive robot-empowered scenario learning influence the learning engagement and motivation of NAS in learning Cantonese? and (2) What is the impact of a robot-empowered scenario learning system on the Cantonese language proficiency of NAS? Fourteen children are invited to participate in a four-day learning program with Boon Boon. The preliminary result indicated that Boon Boon drove students' attention to learning and academic achievement. Future research will focus on long-term assessments of robot-empowered learning's effectiveness and explore the scalability of this approach across diverse educational settings and cultural backgrounds.
MotiBo: The Impact of Interactive Digital Storytelling Robots on Student Motivation through Self-Determination Theory	2026-01-03	Show Creativity is increasingly recognized as an important skill in education, and storytelling can enhance motivation and engagement among students. However, conventional storytelling methods often lack the interactive elements necessary to engage students. To this end, this study examines the impact of an interactive digital storytelling system incorporating a human-like robot on student engagement and creativity. The study aims to compare engagement levels across three modalities: paper-based, PowerPoint, and robot-assisted storytelling, MotiBo. Utilizing a quasi-experimental design, this work involves three groups of students who interact with the storytelling system over a five-day learning. Findings reveal that students using MotiBo exhibit statistically significant improvement in behavioural and cognitive engagement compared to those using traditional methods. These results suggest that the integration of novel technologies can effectively enhance the learning experience, ultimately promoting creativity and self-learning ability in educational settings. Future research will investigate the long-term effects of these technologies on learning outcomes and explore their potential for broader applications in diverse educational contexts.
Stochastic Online Optimization for Cyber-Physical and Robotic Systems	2026-01-03	Show We propose a novel gradient-based online optimization framework for solving stochastic programming problems that frequently arise in the context of cyber-physical and robotic systems. Our problem formulation accommodates constraints that model the evolution of a cyber-physical system, which has, in general, a continuous state and action space, is nonlinear, and where the state is only partially observed. We also incorporate an approximate model of the dynamics as prior knowledge into the learning process and show that even rough estimates of the dynamics can significantly improve the convergence of our algorithms. Our online optimization framework encompasses both gradient descent and quasi-Newton methods, and we provide a unified convergence analysis of our algorithms in a non-convex setting. We also characterize the impact of modeling errors in the system dynamics on the convergence rate of the algorithms. Finally, we evaluate our algorithms in simulations of a flexible beam, a four-legged walking robot, and in real-world experiments with a ping-pong playing robot.	46 pages, 16 figures
SPARC: Spine with Prismatic and Revolute Compliance for Quadruped Robots	2026-01-03	Show We present SPARC, a compact, open-source 3-DoF sagittal-plane spine module that combines revolute (pitch) and prismatic (axial) motion with programmable task-space impedance for quadruped robots. The system integrates three torque-controlled actuators, a 1kHz control board, and protected power electronics in a 1.26kg package. A floating-base impedance controller with dynamics compensation renders tunable spring-damper behavior along horizontal, vertical, and rotational axes. Benchtop experiments validate the approach: quasi-static tests demonstrate linear force-displacement characteristics with commanded horizontal stiffness spanning 300--700~N/m ($\leq 1.5%$ error, $R^2 \geq 0.992$), while dynamic trials confirm predictable damping behavior across multiple settings. Furthermore, simulation studies reveal that adapting spine stiffness based on terrain slope and stride length significantly reduces the cost of transport. SPARC provides an open-source platform for systematic studies of spine compliance in legged locomotion, with complete hardware and firmware resources available for community use. Github repository: https://github.com/YueWang996/sparc.git
EduSim-LLM: An Educational Platform Integrating Large Language Models and Robotic Simulation for Beginners	2026-01-03	Show In recent years, the rapid development of Large Language Models (LLMs) has significantly enhanced natural language understanding and human-computer interaction, creating new opportunities in the field of robotics. However, the integration of natural language understanding into robotic control is an important challenge in the rapid development of human-robot interaction and intelligent automation industries. This challenge hinders intuitive human control over complex robotic systems, limiting their educational and practical accessibility. To address this, we present the EduSim-LLM, an educational platform that integrates LLMs with robot simulation and constructs a language-drive control model that translates natural language instructions into executable robot behavior sequences in CoppeliaSim. We design two human-robot interaction models: direct control and autonomous control, conduct systematic simulations based on multiple language models, and evaluate multi-robot collaboration, motion planning, and manipulation capabilities. Experiential results show that LLMs can reliably convert natural language into structured robot actions; after applying prompt-engineering templates instruction-parsing accuracy improves significantly; as task complexity increases, overall accuracy rate exceeds 88.9% in the highest complexity tests.
RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics	2026-01-03	Show Spatial referring is a fundamental capability of embodied robots to interact with the 3D physical world. However, even with the powerful pretrained vision language models (VLMs), recent approaches are still not qualified to accurately understand the complex 3D scenes and dynamically reason about the instruction-indicated locations for interaction. To this end, we propose RoboRefer, a 3D-aware VLM that can first achieve precise spatial understanding by integrating a disentangled but dedicated depth encoder via supervised fine-tuning (SFT). Moreover, RoboRefer advances generalized multi-step spatial reasoning via reinforcement fine-tuning (RFT), with metric-sensitive process reward functions tailored for spatial referring tasks. To support SFT and RFT training, we introduce RefSpatial, a large-scale dataset of 20M QA pairs (2x prior), covering 31 spatial relations (vs. 15 prior) and supporting complex reasoning processes (up to 5 steps). In addition, we introduce RefSpatial-Bench, a challenging benchmark filling the gap in evaluating spatial referring with multi-step reasoning. Experiments show that SFT-trained RoboRefer achieves state-of-the-art spatial understanding, with an average success rate of 89.6%. RFT-trained RoboRefer further outperforms all other baselines by a large margin, even surpassing Gemini-2.5-Pro by 17.4% in average accuracy on RefSpatial-Bench. Notably, RoboRefer can be integrated with various control policies to execute long-horizon, dynamic tasks across diverse robots (e,g., UR5, G1 humanoid) in cluttered real-world scenes.	Accep... Accepted by NeurIPS 2025. Project page: https://zhoues.github.io/RoboRefer/
RAD: A Dataset and Benchmark for Real-Life Anomaly Detection with Robotic Observations	2026-01-02	Show Anomaly detection is a core capability for robotic perception and industrial inspection, yet most existing benchmarks are collected under controlled conditions with fixed viewpoints and stable illumination, failing to reflect real deployment scenarios. We introduce RAD (Realistic Anomaly Detection), a robot-captured, multi-view dataset designed to stress pose variation, reflective materials, and viewpoint-dependent defect visibility. RAD covers 13 everyday object categories and four realistic defect types--scratched, missing, stained, and squeezed--captured from over 60 robot viewpoints per object under uncontrolled lighting. We benchmark a wide range of state-of-the-art approaches, including 2D feature-based methods, 3D reconstruction pipelines, and vision-language models (VLMs), under a pose-agnostic setting. Surprisingly, we find that mature 2D feature-embedding methods consistently outperform recent 3D and VLM-based approaches at the image level, while the performance gap narrows for pixel-level localization. Our analysis reveals that reflective surfaces, geometric symmetry, and sparse viewpoint coverage fundamentally limit current geometry-based and zero-shot methods. RAD establishes a challenging and realistic benchmark for robotic anomaly detection, highlighting critical open problems beyond controlled laboratory settings.
From Perception to Symbolic Task Planning: Vision-Language Guided Human-Robot Collaborative Structured Assembly	2026-01-02	Show Human-robot collaboration (HRC) in structured assembly requires reliable state estimation and adaptive task planning under noisy perception and human interventions. To address these challenges, we introduce a design-grounded human-aware planning framework for human-robot collaborative structured assembly. The framework comprises two coupled modules. Module I, Perception-to-Symbolic State (PSS), employs vision-language models (VLMs) based agents to align RGB-D observations with design specifications and domain knowledge, synthesizing verifiable symbolic assembly states. It outputs validated installed and uninstalled component sets for online state tracking. Module II, Human-Aware Planning and Replanning (HPR), performs task-level multi-robot assignment and updates the plan only when the observed state deviates from the expected execution outcome. It applies a minimal-change replanning rule to selectively revise task assignments and preserve plan stability even under human interventions. We validate the framework on a 27-component timber-frame assembly. The PSS module achieves 97% state synthesis accuracy, and the HPR module maintains feasible task progression across diverse HRC scenarios. Results indicate that integrating VLM-based perception with knowledge-driven planning improves robustness of state estimation and task planning under dynamic conditions.
Calling for Backup: How Children Navigate Successive Robot Communication Failures	2026-01-02	Show How do children respond to repeated robot errors? While prior research has examined adult reactions to successive robot errors, children's responses remain largely unexplored. In this study, we explore children's reactions to robot social errors and performance errors. For the latter, this study reproduces the successive robot failure paradigm of Liu et al. with child participants (N=59, ages 8-10) to examine how young users respond to repeated robot conversational errors. Participants interacted with a robot that failed to understand their prompts three times in succession, with their behavioral responses video-recorded and analyzed. We found both similarities and differences compared to adult responses from the original study. Like adults, children adjusted their prompts, modified their verbal tone, and exhibited increasingly emotional non-verbal responses throughout successive errors. However, children demonstrated more disengagement behaviors, including temporarily ignoring the robot or actively seeking an adult. Errors did not affect participants' perception of the robot, suggesting more flexible conversational expectations in children. These findings inform the design of more effective and developmentally appropriate human-robot interaction systems for young users.
RoboReward: General-Purpose Vision-Language Reward Models for Robotics	2026-01-02	Show A well-designed reward is critical for effective reinforcement learning-based policy improvement. In real-world robotic domains, obtaining such rewards typically requires either labor-intensive human labeling or brittle, handcrafted objectives. Vision-language models (VLMs) have shown promise as automatic reward models, yet their effectiveness on real robot tasks is poorly understood. In this work, we aim to close this gap by introducing (1) \textbf{RoboReward}, a robotics reward dataset and benchmark built on large-scale real-robot corpora from Open X-Embodiment (OXE) and RoboArena, and (2) vision-language reward models trained on this dataset (RoboReward 4B/8B). Because OXE is success-heavy and lacks failure examples, we propose a \emph{negative examples data augmentation} pipeline that generates calibrated \emph{negatives} and \emph{near-misses} via counterfactual relabeling of successful episodes and temporal clipping to create partial-progress outcomes from the same videos. Using this framework, we produce an extensive training and evaluation dataset that spans diverse tasks and embodiments and enables systematic evaluation of whether state-of-the-art VLMs can reliably provide rewards for robotics. Our evaluation of leading open-weight and proprietary VLMs reveals that no model excels across all tasks, underscoring substantial room for improvement. We then train general-purpose 4B- and 8B-parameter models that outperform much larger VLMs in assigning rewards for short-horizon robotic tasks. Finally, we deploy the 8B-parameter reward VLM in real-robot reinforcement learning and find that it improves policy learning over Gemini Robotics-ER 1.5, a frontier physical reasoning VLM trained on robotics data, by a large margin, while substantially narrowing the gap to RL training with human-provided rewards.
Iterative Tuning of Nonlinear Model Predictive Control for Robotic Manufacturing Tasks	2026-01-02	Show Manufacturing processes are often perturbed by drifts in the environment and wear in the system, requiring control re-tuning even in the presence of repetitive operations. This paper presents an iterative learning framework for automatic tuning of Nonlinear Model Predictive Control (NMPC) weighting matrices based on task-level performance feedback. Inspired by norm-optimal Iterative Learning Control (ILC), the proposed method adaptively adjusts NMPC weights Q and R across task repetitions to minimize key performance indicators (KPIs) related to tracking accuracy, control effort, and saturation. Unlike gradient-based approaches that require differentiating through the NMPC solver, we construct an empirical sensitivity matrix, enabling structured weight updates without analytic derivatives. The framework is validated through simulation on a UR10e robot performing carbon fiber winding on a tetrahedral core. Results demonstrate that the proposed approach converges to near-optimal tracking performance (RMSE within 0.3% of offline Bayesian Optimization (BO)) in just 4 online repetitions, compared to 100 offline evaluations required by BO algorithm. The method offers a practical solution for adaptive NMPC tuning in repetitive robotic tasks, combining the precision of carefully optimized controllers with the flexibility of online adaptation.
Vision-based Goal-Reaching Control for Mobile Robots Using a Hierarchical Learning Framework	2026-01-02	Show Reinforcement learning (RL) is effective in many robotic applications, but it requires extensive exploration of the state-action space, during which behaviors can be unsafe. This significantly limits its applicability to large robots with complex actuators operating on unstable terrain. Hence, to design a safe goal-reaching control framework for large-scale robots, this paper decomposes the whole system into a set of tightly coupled functional modules. 1) A real-time visual pose estimation approach is employed to provide accurate robot states to 2) an RL motion planner for goal-reaching tasks that explicitly respects robot specifications. The RL module generates real-time smooth motion commands for the actuator system, independent of its underlying dynamic complexity. 3) In the actuation mechanism, a supervised deep learning model is trained to capture the complex dynamics of the robot and provide this model to 4) a model-based robust adaptive controller that guarantees the wheels track the RL motion commands even on slip-prone terrain. 5) Finally, to reduce human intervention, a mathematical safety supervisor monitors the robot, stops it on unsafe faults, and autonomously guides it back to a safe inspection area. The proposed framework guarantees uniform exponential stability of the actuation system and safety of the whole operation. Experiments on a 6,000 kg robot in different scenarios confirm the effectiveness of the proposed framework.
NMPC-Augmented Visual Navigation and Safe Learning Control for Large-Scale Mobile Robots	2026-01-02	Show A large-scale mobile robot (LSMR) is a high-order multibody system that often operates on loose, unconsolidated terrain, which reduces traction. This paper presents a comprehensive navigation and control framework for an LSMR that ensures stability and safety-defined performance, delivering robust operation on slip-prone terrain by jointly leveraging high-performance techniques. The proposed architecture comprises four main modules: (1) a visual pose-estimation module that fuses onboard sensors and stereo cameras to provide an accurate, low-latency robot pose, (2) a high-level nonlinear model predictive control that updates the wheel motion commands to correct robot drift from the robot reference pose on slip-prone terrain, (3) a low-level deep neural network control policy that approximates the complex behavior of the wheel-driven actuation mechanism in LSMRs, augmented with robust adaptive control to handle out-of-distribution disturbances, ensuring that the wheels accurately track the updated commands issued by high-level control module, and (4) a logarithmic safety module to monitor the entire robot stack and guarantees safe operation. The proposed low-level control framework guarantees uniform exponential stability of the actuation subsystem, while the safety module ensures the whole system-level safety during operation. Comparative experiments on a 6,000 kg LSMR actuated by two complex electro-hydrostatic drives, while synchronizing modules operating at different frequencies.
Digital Twin based Automatic Reconfiguration of Robotic Systems in Smart Environments	2026-01-02	Show Robotic systems have become integral to smart environments, enabling applications ranging from urban surveillance and automated agriculture to industrial automation. However, their effective operation in dynamic settings - such as smart cities and precision farming - is challenged by continuously evolving topographies and environmental conditions. Traditional control systems often struggle to adapt quickly, leading to inefficiencies or operational failures. To address this limitation, we propose a novel framework for autonomous and dynamic reconfiguration of robotic controllers using Digital Twin technology. Our approach leverages a virtual replica of the robot's operational environment to simulate and optimize movement trajectories in response to real-world changes. By recalculating paths and control parameters in the Digital Twin and deploying the updated code to the physical robot, our method ensures rapid and reliable adaptation without manual intervention. This work advances the integration of Digital Twins in robotics, offering a scalable solution for enhancing autonomy in smart, dynamic environments.	Accep... Accepted for presentation to 11th IEEE International Smart Cities Conference (ISC2 2025)
Tackling the Kidnapped Robot Problem via Sparse Feasible Hypothesis Sampling and Reliable Batched Multi-Stage Inference	2026-01-02	Show This paper addresses the Kidnapped Robot Problem (KRP), a core localization challenge of relocalizing a robot in a known map without prior pose estimate when localization loss or at SLAM initialization. For this purpose, a passive 2-D global relocalization framework is proposed. It estimates the global pose efficiently and reliably from a single LiDAR scan and an occupancy grid map while the robot remains stationary, thereby enhancing the long-term autonomy of mobile robots. The proposed framework casts global relocalization as a non-convex problem and solves it via the multi-hypothesis scheme with batched multi-stage inference and early termination, balancing completeness and efficiency. The Rapidly-exploring Random Tree (RRT), under traversability constraints, asymptotically covers the reachable space to generate sparse, uniformly distributed feasible positional hypotheses, fundamentally reducing the sampling space. The hypotheses are preliminarily ordered by the proposed Scan Mean Absolute Difference (SMAD), a coarse beam-error level metric that facilitates the early termination by prioritizing high-likelihood candidates. The SMAD computation is optimized for non-panoramic scans. The Translation-Affinity Scan-to-Map Alignment Metric (TAM) is proposed for reliable orientation selection at hypothesized positions and accurate final pose evaluation to mitigate degradation in conventional likelihood-field metrics under translational uncertainty induced by sparse hypotheses, as well as non-panoramic LiDAR scan and environmental changes. Real-world experiments on a resource-constrained mobile robot with non-panoramic LiDAR scans show that the proposed framework achieves competitive performance in both global relocalization success rate and computational efficiency.	10 pa... 10 pages, 8 figures. This work has been submitted to the IEEE for possible publication

Vision Language Model

Title	Date	Abstract	Comment
Musical Score Understanding Benchmark: Evaluating Large Language Models' Comprehension of Complete Musical Scores	2026-01-06	Show Understanding complete musical scores entails integrated reasoning over pitch, rhythm, harmony, and large-scale structure, yet the ability of Large Language Models and Vision-Language Models to interpret full musical notation remains insufficiently examined. We introduce the Musical Score Understanding Benchmark (MSU-Bench), the first large-scale, human-curated benchmark for score-level musical understanding across textual (ABC notation) and visual (PDF) modalities. MSU-Bench contains 1,800 generative Question-Answering pairs from works by Bach, Beethoven, Chopin, Debussy, and others, organised into four levels of increasing difficulty, ranging from onset information to texture and form. Evaluations of more than fifteen state-of-the-art models, in both zero-shot and fine-tuned settings, reveal pronounced modality gaps, unstable level-wise performance, and challenges in maintaining multilevel correctness. Fine-tuning substantially improves results across modalities while preserving general knowledge, positioning MSU-Bench as a robust foundation for future research in multimodal reasoning. To facilitate further research, we publicly release MSU-Bench and all associated resources.
ReCCur: A Recursive Corner-Case Curation Framework for Robust Vision-Language Understanding in Open and Edge Scenarios	2026-01-06	Show Corner cases are rare or extreme scenarios that drive real-world failures, but they are difficult to curate at scale: web data are noisy, labels are brittle, and edge deployments preclude large retraining. We present ReCCur (Recursive Corner-Case Curation), a low-compute framework that converts noisy web imagery into auditable fine-grained labels via a multi-agent recursive pipeline. First, large-scale data acquisition and filtering expands a domain vocabulary with a vision-language model (VLM), crawls the web, and enforces tri-modal (image, description, keyword) consistency with light human spot checks to yield refined candidates. Next, mixture-of-experts knowledge distillation uses complementary encoders (e.g., CLIP, DINOv2, BEiT) for kNN voting with dual-confidence activation and uncertainty sampling, converging to a high-precision set. Finally, region-evidence VLM adversarial labeling pairs a proposer (multi-granularity regions and semantic cues) with a validator (global and local chained consistency) to produce explainable labels and close the loop. On realistic corner-case scenarios (e.g., flooded-car inspection), ReCCur runs on consumer-grade GPUs, steadily improves purity and separability, and requires minimal human supervision, providing a practical substrate for downstream training and evaluation under resource constraints. Code and dataset will be released.
Zoom-IQA: Image Quality Assessment with Reliable Region-Aware Reasoning	2026-01-06	Show Image Quality Assessment (IQA) is a long-standing problem in computer vision. Previous methods typically focus on predicting numerical scores without explanation or provide low-level descriptions lacking precise scores. Recent reasoning-based vision language models (VLMs) have shown strong potential for IQA, enabling joint generation of quality descriptions and scores. However, we notice that existing VLM-based IQA methods tend to exhibit unreliable reasoning due to their limited capability of integrating visual and textual cues. In this work, we introduce Zoom-IQA, a VLM-based IQA model to explicitly emulate key cognitive behaviors: uncertainty awareness, region reasoning, and iterative refinement. Specifically, we present a two-stage training pipeline: 1) supervised fine-tuning (SFT) on our Grounded-Rationale-IQA (GR-IQA) dataset to teach the model to ground its assessments in key regions; and 2) reinforcement learning (RL) for dynamic policy exploration, primarily stabilized by our KL-Coverage regularizer to prevent reasoning and scoring diversity collapse, and supported by a Progressive Re-sampling Strategy to mitigate annotation bias. Extensive experiments show that Zoom-IQA achieves improved robustness, explainability, and generalization. The application to downstream tasks, such as image restoration, further demonstrates the effectiveness of Zoom-IQA.	Proje... Project Page: https://ethanliang99.github.io/ZOOMIQA-Projectpage
CaTS-Bench: Can Language Models Describe Time Series?	2026-01-06	Show Time series captioning, the task of describing time series in natural language, requires numeric and temporal reasoning, trend interpretation, and contextual understanding. Existing benchmarks, however, often rely on fully synthetic or generic captions, and typically neglect metadata and visual representations. We introduce \textbf{CaTS-Bench}, a comprehensive benchmark for \textbf{C}ontext-\textbf{a}ware \textbf{T}ime \textbf{S}eries reasoning across $11$ diverse domains, centered on a gold-standard evaluation set of $1746$ human-rewritten captions that measure how effectively models translate numeric trends into immediately interpretable narratives. To address the scarcity of human-annotated data, we also propose a scalable pipeline for generating high-fidelity synthetic captions, the quality of which we validate. We evaluate leading Vision-Language Models on our benchmark, revealing that even proprietary models struggle to capture numeric nuances in temporal descriptions, while finetuning open-source models on synthetic data yields substantial performance gains. Finally, release a diagnostic suite of $910$ multiple-choice questions and tailored numeric metrics to gauge time-series-specific reasoning capabilities, establishing CaTS-Bench as a reliable foundation for grounded, multimodal language generation in numeric domains.	8 pag... 8 pages, 6 figures, 3 tables in the main paper. Many more in the appendix
RPIQ: Residual-Projected Multi-Collaboration Closed-Loop and Single Instance Quantization for Visually Impaired Assistance	2026-01-06	Show Visually impaired users face significant challenges in daily information access and real-time environmental perception, and there is an urgent need for intelligent assistive systems with accurate recognition capabilities. Although large-scale models provide effective solutions for perception and reasoning, their practical deployment on assistive devices is severely constrained by excessive memory consumption and high inference costs. Moreover, existing quantization strategies often ignore inter-block error accumulation, leading to degraded model stability. To address these challenges, this study proposes a novel quantization framework -- Residual-Projected Multi-Collaboration Closed-Loop and Single Instance Quantization(RPIQ), whose quantization process adopts a multi-collaborative closed-loop compensation scheme based on Single Instance Calibration and Gauss-Seidel Iterative Quantization. Experiments on various types of large-scale models, including language models such as OPT, Qwen, and LLaMA, as well as vision-language models such as CogVLM2, demonstrate that RPIQ can compress models to 4-bit representation while significantly reducing peak memory consumption (approximately 60%-75% reduction compared to original full-precision models). The method maintains performance highly close to full-precision models across multiple language and visual tasks, and exhibits excellent recognition and reasoning capabilities in key applications such as text understanding and visual question answering in complex scenarios. While verifying the effectiveness of RPIQ for deployment in real assistive systems, this study also advances the computational efficiency and reliability of large models, enabling them to provide visually impaired users with the required information accurately and rapidly.
Scene-Aware Vectorized Memory Multi-Agent Framework with Cross-Modal Differentiated Quantization VLMs for Visually Impaired Assistance	2026-01-06	Show Visually impaired individuals face significant challenges in environmental perception. Traditional assistive technologies often lack adaptive intelligence, focusing on individual components rather than integrated systems. While Vision-Language Models (VLMs) offer a promising path to richer, integrated understanding, their deployment is severely limited by substantial computational requirements, demanding dozens of gigabytes of memory. To address these gaps in computational efficiency and integrated design, this study proposes a dual technological innovation framework: a cross-modal differentiated quantization framework for VLMs and a scene-aware vectorized memory multi-agent system. The quantization framework implements differentiated strategies, reducing memory from 38GB to 11.3GB. The multi-agent system uses vectorized memory and perception-memory-reasoning workflows to provide environmental information beyond the current view, achieving 2.83-3.52s latency to initial speech output. Experiments show the quantized 19B-parameter model only experiences a 2.05% performance drop on MMBench and maintains 63.7 accuracy on OCR-VQA (original: 64.9), outperforming smaller models with equivalent memory. This research advances computational efficiency and assistive technology, offering comprehensive assistance in scene perception, text recognition, and navigation.	28 pages,9 figures
FCMBench: A Comprehensive Financial Credit Multimodal Benchmark for Real-world Applications	2026-01-06	Show As multimodal AI becomes widely used for credit risk assessment and document review, a domain-specific benchmark is urgently needed that (1) reflects documents and workflows specific to financial credit applications, (2) includes credit-specific understanding and real-world robustness, and (3) preserves privacy compliance without sacrificing practical utility. Here, we introduce FCMBench-V1.0 -- a large-scale financial credit multimodal benchmark for real-world applications, covering 18 core certificate types, with 4,043 privacy-compliant images and 8,446 QA samples. The FCMBench evaluation framework consists of three dimensions: Perception, Reasoning, and Robustness, including 3 foundational perception tasks, 4 credit-specific reasoning tasks that require decision-oriented understanding of visual evidence, and 10 real-world acquisition artifact types for robustness stress testing. To reconcile compliance with realism, we construct all samples via a closed synthesis-capture pipeline: we manually synthesize document templates with virtual content and capture scenario-aware images in-house. This design also mitigates pre-training data leakage by avoiding web-sourced or publicly released images. FCMBench can effectively discriminate performance disparities and robustness across modern vision-language models. Extensive experiments were conducted on 23 state-of-the-art vision-language models (VLMs) from 14 top AI companies and research institutes. Among them, Gemini 3 Pro achieves the best F1(%) score as a commercial model (64.61), Qwen3-VL-235B achieves the best score as an open-source baseline (57.27), and our financial credit-specific model, Qfin-VL-Instruct, achieves the top overall score (64.92). Robustness evaluations show that even top-performing models suffer noticeable performance drops under acquisition artifacts.
Intervene-All-Paths: Unified Mitigation of LVLM Hallucinations across Alignment Formats	2026-01-06	Show Despite their impressive performance across a wide range of tasks, Large Vision-Language Models (LVLMs) remain prone to hallucination. In this study, we propose a comprehensive intervention framework aligned with the transformer's causal architecture in LVLMs, integrating the effects of different intervention paths on hallucination. We find that hallucinations in LVLMs do not arise from a single causal path, but rather from the interplay among image-to-input-text, image-to-output-text, and text-to-text pathways. For the first time, we also find that LVLMs rely on different pathways depending on the question-answer alignment format. Building on these insights, we propose simple yet effective methods to identify and intervene on critical hallucination heads within each pathway, tailored to discriminative and generative formats. Experiments across multiple benchmarks demonstrate that our approach consistently reduces hallucinations across diverse alignment types.	Accep... Accepted to NeurIPS 2025, Project Page: https://github.com/SooLab/AllPath
RoboTracer: Mastering Spatial Trace with Reasoning in Vision-Language Models for Robotics	2026-01-06	Show Spatial tracing, as a fundamental embodied interaction ability for robots, is inherently challenging as it requires multi-step metric-grounded reasoning compounded with complex spatial referring and real-world metric measurement. However, existing methods struggle with this compositional task. To this end, we propose RoboTracer, a 3D-aware VLM that first achieves both 3D spatial referring and measuring via a universal spatial encoder and a regression-supervised decoder to enhance scale awareness during supervised fine-tuning (SFT). Moreover, RoboTracer advances multi-step metric-grounded reasoning via reinforcement fine-tuning (RFT) with metric-sensitive process rewards, supervising key intermediate perceptual cues to accurately generate spatial traces. To support SFT and RFT training, we introduce TraceSpatial, a large-scale dataset of 30M QA pairs, spanning outdoor/indoor/tabletop scenes and supporting complex reasoning processes (up to 9 steps). We further present TraceSpatial-Bench, a challenging benchmark filling the gap to evaluate spatial tracing. Experimental results show that RoboTracer surpasses baselines in spatial understanding, measuring, and referring, with an average success rate of 79.1%, and also achieves SOTA performance on TraceSpatial-Bench by a large margin, exceeding Gemini-2.5-Pro by 36% accuracy. Notably, RoboTracer can be integrated with various control policies to execute long-horizon, dynamic tasks across diverse robots (UR5, G1 humanoid) in cluttered real-world scenes. See the project page at https://zhoues.github.io/RoboTracer.	Proje... Project page: https://zhoues.github.io/RoboTracer
DarkEQA: Benchmarking Vision-Language Models for Embodied Question Answering in Low-Light Indoor Environments	2026-01-06	Show Vision Language Models (VLMs) are increasingly adopted as central reasoning modules for embodied agents. Existing benchmarks evaluate their capabilities under ideal, well-lit conditions, yet robust 24/7 operation demands performance under a wide range of visual degradations, including low-light conditions at night or in dark environments--a core necessity that has been largely overlooked. To address this underexplored challenge, we present DarkEQA, an open-source benchmark for evaluating EQA-relevant perceptual primitives under multi-level low-light conditions. DarkEQA isolates the perception bottleneck by evaluating question answering from egocentric observations under controlled degradations, enabling attributable robustness analysis. A key design feature of DarkEQA is its physical fidelity: visual degradations are modeled in linear RAW space, simulating physics-based illumination drop and sensor noise followed by an ISP-inspired rendering pipeline. We demonstrate the utility of DarkEQA by evaluating a wide range of state-of-the-art VLMs and Low-Light Image Enhancement (LLIE) models. Our analysis systematically reveals VLMs' limitations when operating under these challenging visual conditions. Project website: https://darkeqa-benchmark.github.io/	Submi... Submitted to IEEE Robotics and Automation Letters (RA-L)
MIRAGE: A Benchmark for Multimodal Information-Seeking and Reasoning in Agricultural Expert-Guided Conversations	2026-01-06	Show We introduce MIRAGE, a new benchmark for multimodal expert-level reasoning and decision-making in consultative interaction settings. Designed for the agriculture domain, MIRAGE captures the full complexity of expert consultations by combining natural user queries, expert-authored responses, and image-based context, offering a high-fidelity benchmark for evaluating models on grounded reasoning, clarification strategies, and long-form generation in a real-world, knowledge-intensive domain. Grounded in over 35,000 real user-expert interactions and curated through a carefully designed multi-step pipeline, MIRAGE spans diverse crop health, pest diagnosis, and crop management scenarios. The benchmark includes more than 7,000 unique biological entities, covering plant species, pests, and diseases, making it one of the most taxonomically diverse benchmarks available for vision-language models, grounded in the real world. Unlike existing benchmarks that rely on well-specified user inputs and closed-set taxonomies, MIRAGE features underspecified, context-rich scenarios with open-world settings, requiring models to infer latent knowledge gaps, handle rare entities, and either proactively guide the interaction or respond. Project Page: https://mirage-benchmark.github.io	Accep... Accepted to NeurIPS 2025
Aha Moment Revisited: Are VLMs Truly Capable of Self Verification in Inference-time Scaling?	2026-01-06	Show Inference time techniques such as decoding time scaling and self refinement have been shown to substantially improve mathematical reasoning in large language models (LLMs), largely attributed to emergent self correction and self verification behaviors often elicited through reinforcement learning (RL). In this work, we ask whether the same recipe transfers to vision language models (VLMs), especially RL finetuned variants that claim strong visual mathematical reasoning. Through extensive evaluation, we reach three main findings that differ markedly from text only models. First, generation time capability matters more than verification and refinement: simple majority voting consistently and substantially outperforms verification centric strategies such as best of N with self verification. Second, behaviors often associated with RL tuned models at inference time, such as the 'Aha moment,' do not yield reliable reasoning performance improvements. Third, visual information is not effectively integrated into the model's self verification process. Overall, our analysis highlights a key limitation: current RL trained VLMs derive limited benefit from self verification in the visual modality, which constrains the effectiveness of inference time scaling for visual mathematical reasoning.	Neuri... Neurips 2025 Multimodal Algorithmic Reasoning Workshop Oral. In submission
RxnCaption: Reformulating Reaction Diagram Parsing as Visual Prompt Guided Captioning	2026-01-06	Show Large-scale chemical reaction datasets are crucial for AI research in chemistry. However, existing chemical reaction data often exist as images within papers, making them not machine-readable and unusable for training machine learning models. In response to this challenge, we propose the RxnCaption framework for the task of chemical Reaction Diagram Parsing (RxnDP). Our framework reformulates the traditional coordinate prediction driven parsing process into an image captioning problem, which Large Vision Language Models (LVLMs) handle naturally. We introduce a strategy termed BBox and Index as Visual Prompt (BIVP), which uses our state-of-the-art molecular detector, MolYOLO, to pre-draw molecular bounding boxes and indices directly onto the input image. This turns the downstream parsing into a natural-language description problem. Extensive experiments show that the BIVP strategy significantly improves structural extraction quality while simplifying model design. We further construct the RxnCaption-15k dataset, an order of magnitude larger than prior real-world literature benchmarks, with a balanced test subset across four layout archetypes. Experiments demonstrate that RxnCaption-VL achieves state-of-the-art performance on multiple metrics. We believe our method, dataset, and models will advance structured information extraction from chemical literature and catalyze broader AI applications in chemistry. We will release data, models, and code on GitHub.
Agentic Physical AI toward a Domain-Specific Foundation Model for Nuclear Reactor Control	2026-01-06	Show The prevailing paradigm in AI for physical systems, scaling general-purpose foundation models toward universal multimodal reasoning, confronts a fundamental barrier at the control interface. Recent benchmarks show that even frontier vision-language models achieve only 50-53% accuracy on basic quantitative physics tasks, behaving as approximate guessers that preserve semantic plausibility while violating physical constraints. This input unfaithfulness is not a scaling deficiency but a structural limitation. Perception-centric architectures optimize parameter-space imitation, whereas safety-critical control demands outcome-space guarantees over executed actions. Here, we present a fundamentally different pathway toward domain-specific foundation models by introducing compact language models operating as Agentic Physical AI, in which policy optimization is driven by physics-based validation rather than perceptual inference. We train a 360-million-parameter model on synthetic reactor control scenarios, scaling the dataset from 10^3 to 10^5 examples. This induces a sharp phase transition absent in general-purpose models. Small-scale systems exhibit high-variance imitation with catastrophic tail risk, while large-scale models undergo variance collapse exceeding 500x reduction, stabilizing execution-level behavior. Despite balanced exposure to four actuation families, the model autonomously rejects approximately 70% of the training distribution and concentrates 95% of runtime execution on a single-bank strategy. Learned representations transfer across distinct physics and continuous input modalities without architectural modification.
VINO: A Unified Visual Generator with Interleaved OmniModal Context	2026-01-05	Show We present VINO, a unified visual generator that performs image and video generation and editing within a single framework. Instead of relying on task-specific models or independent modules for each modality, VINO uses a shared diffusion backbone that conditions on text, images and videos, enabling a broad range of visual creation and editing tasks under one model. Specifically, VINO couples a vision-language model (VLM) with a Multimodal Diffusion Transformer (MMDiT), where multimodal inputs are encoded as interleaved conditioning tokens, and then used to guide the diffusion process. This design supports multi-reference grounding, long-form instruction following, and coherent identity preservation across static and dynamic content, while avoiding modality-specific architectural components. To train such a unified system, we introduce a multi-stage training pipeline that progressively expands a video generation base model into a unified, multi-task generator capable of both image and video input and output. Across diverse generation and editing benchmarks, VINO demonstrates strong visual quality, faithful instruction following, improved reference and attribute preservation, and more controllable multi-identity edits. Our results highlight a practical path toward scalable unified visual generation, and the promise of interleaved, in-context computation as a foundation for general-purpose visual creation.	Proje... Project page: https://sotamak1r.github.io/VINO-web/
Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach	2026-01-05	Show Recent advances in vision-language models have opened up new possibilities for reasoning-driven image geolocalization. However, existing approaches often rely on synthetic reasoning annotations or external image retrieval, which can limit interpretability and generalizability. In this paper, we present Geo-R, a retrieval-free framework that uncovers structured reasoning paths from existing ground-truth coordinates and optimizes geolocation accuracy via reinforcement learning. We propose the Chain of Region, a rule-based hierarchical reasoning paradigm that generates precise, interpretable supervision by mapping GPS coordinates to geographic entities (e.g., country, province, city) without relying on model-generated or synthetic labels. Building on this, we introduce a lightweight reinforcement learning strategy with coordinate-aligned rewards based on Haversine distance, enabling the model to refine predictions through spatially meaningful feedback. Our approach bridges structured geographic reasoning with direct spatial supervision, yielding improved localization accuracy, stronger generalization, and more transparent inference. Experimental results across multiple benchmarks confirm the effectiveness of Geo-R, establishing a new retrieval-free paradigm for scalable and interpretable image geolocalization. To facilitate further research and ensure reproducibility, both the model and code will be made publicly available.	Accep... Accepted to AAAI 2026. Project Page: https://github.com/aialt/geo-r
DatBench: Discriminative, Faithful, and Efficient VLM Evaluations	2026-01-05	Show Empirical evaluation serves as the primary compass guiding research progress in foundation models. Despite a large body of work focused on training frontier vision-language models (VLMs), approaches to their evaluation remain nascent. To guide their maturation, we propose three desiderata that evaluations should satisfy: (1) faithfulness to the modality and application, (2) discriminability between models of varying quality, and (3) efficiency in compute. Through this lens, we identify critical failure modes that violate faithfulness and discriminability, misrepresenting model capabilities: (i) multiple-choice formats reward guessing, poorly reflect downstream use cases, and saturate early as models improve; (ii) blindly solvable questions, which can be answered without images, constitute up to 70% of some evaluations; and (iii) mislabeled or ambiguous samples compromise up to 42% of examples in certain datasets. Regarding efficiency, the computational burden of evaluating frontier models has become prohibitive: by some accounts, nearly 20% of development compute is devoted to evaluation alone. Rather than discarding existing benchmarks, we curate them via transformation and filtering to maximize fidelity and discriminability. We find that converting multiple-choice questions to generative tasks reveals sharp capability drops of up to 35%. In addition, filtering blindly solvable and mislabeled samples improves discriminative power while simultaneously reducing computational cost. We release DatBench-Full, a cleaned evaluation suite of 33 datasets spanning nine VLM capabilities, and DatBench, a discriminative subset that achieves 13x average speedup (up to 50x) while closely matching the discriminative power of the original datasets. Our work outlines a path toward evaluation practices that are both rigorous and sustainable as VLMs continue to scale.
Foundation models on the bridge: Semantic hazard detection and safety maneuvers for maritime autonomy with vision-language models	2026-01-05	Show The draft IMO MASS Code requires autonomous and remotely supervised maritime vessels to detect departures from their operational design domain, enter a predefined fallback that notifies the operator, permit immediate human override, and avoid changing the voyage plan without approval. Meeting these obligations in the alert-to-takeover gap calls for a short-horizon, human-overridable fallback maneuver. Classical maritime autonomy stacks struggle when the correct action depends on meaning (e.g., diver-down flag means people in the water, fire close by means hazard). We argue (i) that vision-language models (VLMs) provide semantic awareness for such out-of-distribution situations, and (ii) that a fast-slow anomaly pipeline with a short-horizon, human-overridable fallback maneuver makes this practical in the handover window. We introduce Semantic Lookout, a camera-only, candidate-constrained VLM fallback maneuver selector that selects one cautious action (or station-keeping) from water-valid, world-anchored trajectories under continuous human authority. On 40 harbor scenes we measure per-call scene understanding and latency, alignment with human consensus (model majority-of-three voting), short-horizon risk-relief on fire hazard scenes, and an on-water alert->fallback maneuver->operator handover. Sub-10 s models retain most of the awareness of slower state-of-the-art models. The fallback maneuver selector outperforms geometry-only baselines and increases standoff distance on fire scenes. A field run verifies end-to-end operation. These results support VLMs as semantic fallback maneuver selectors compatible with the draft IMO MASS Code, within practical latency budgets, and motivate future work on domain-adapted, hybrid autonomy that pairs foundation-model semantics with multi-sensor bird's-eye-view perception and short-horizon replanning. Website: kimachristensen.github.io/bridge_policy	17 pa... 17 pages without bibliography or appendix. The main paper has 16 figures. Paper webpage can be found at https://kimachristensen.github.io/bridge_policy/
BiPrompt: Bilateral Prompt Optimization for Visual and Textual Debiasing in Vision-Language Models	2026-01-05	Show Vision language foundation models such as CLIP exhibit impressive zero-shot generalization yet remain vulnerable to spurious correlations across visual and textual modalities. Existing debiasing approaches often address a single modality either visual or textual leading to partial robustness and unstable adaptation under distribution shifts. We propose a bilateral prompt optimization framework (BiPrompt) that simultaneously mitigates non-causal feature reliance in both modalities during test-time adaptation. On the visual side, it employs structured attention-guided erasure to suppress background activations and enforce orthogonal prediction consistency between causal and spurious regions. On the textual side, it introduces balanced prompt normalization, a learnable re-centering mechanism that aligns class embeddings toward an isotropic semantic space. Together, these modules jointly minimize conditional mutual information between spurious cues and predictions, steering the model toward causal, domain invariant reasoning without retraining or domain supervision. Extensive evaluations on real-world and synthetic bias benchmarks demonstrate consistent improvements in both average and worst-group accuracies over prior test-time debiasing methods, establishing a lightweight yet effective path toward trustworthy and causally grounded vision-language adaptation.	Accep... Accepted at the AAAI 2026 Workshop AIR-FM, Assessing and Improving Reliability of Foundation Models in the Real World
Multimodal Adversarial Defense for Vision-Language Models by Leveraging One-To-Many Relationships	2026-01-05	Show Pre-trained vision-language (VL) models are highly vulnerable to adversarial attacks. However, existing defense methods primarily focus on image classification, overlooking two key aspects of VL tasks: multimodal attacks, where both image and text can be perturbed, and the one-to-many relationship of images and texts, where a single image can correspond to multiple textual descriptions and vice versa (1:N and N:1). This work is the first to explore defense strategies against multimodal attacks in VL tasks, whereas prior VL defense methods focus on vision robustness. We propose multimodal adversarial training (MAT), which incorporates adversarial perturbations in both image and text modalities during training, significantly outperforming existing unimodal defenses. Furthermore, we discover that MAT is limited by deterministic one-to-one (1:1) image-text pairs in VL training data. To address this, we conduct a comprehensive study on leveraging one-to-many relationships to enhance robustness, investigating diverse augmentation techniques. Our analysis shows that, for a more effective defense, augmented image-text pairs should be well-aligned, diverse, yet avoid distribution shift -- conditions overlooked by prior research. This work pioneers defense strategies against multimodal attacks, providing insights for building robust VLMs from both optimization and data perspectives. Our code is publicly available at https://github.com/CyberAgentAILab/multimodal-adversarial-training.	WACV ... WACV 2026 Accepted. Code available at https://github.com/CyberAgentAILab/multimodal-adversarial-training
Genie Sim 3.0 : A High-Fidelity Comprehensive Simulation Platform for Humanoid Robot	2026-01-05	Show The development of robust and generalizable robot learning models is critically contingent upon the availability of large-scale, diverse training data and reliable evaluation benchmarks. Collecting data in the physical world poses prohibitive costs and scalability challenges, and prevailing simulation benchmarks frequently suffer from fragmentation, narrow scope, or insufficient fidelity to enable effective sim-to-real transfer. To address these challenges, we introduce Genie Sim 3.0, a unified simulation platform for robotic manipulation. We present Genie Sim Generator, a large language model (LLM)-powered tool that constructs high-fidelity scenes from natural language instructions. Its principal strength resides in rapid and multi-dimensional generalization, facilitating the synthesis of diverse environments to support scalable data collection and robust policy evaluation. We introduce the first benchmark that pioneers the application of LLM for automated evaluation. It leverages LLM to mass-generate evaluation scenarios and employs Vision-Language Model (VLM) to establish an automated assessment pipeline. We also release an open-source dataset comprising more than 10,000 hours of synthetic data across over 200 tasks. Through systematic experimentation, we validate the robust zero-shot sim-to-real transfer capability of our open-source dataset, demonstrating that synthetic data can server as an effective substitute for real-world data under controlled conditions for scalable policy training. For code and dataset details, please refer to: https://github.com/AgibotTech/genie_sim.
Agentic Retoucher for Text-To-Image Generation	2026-01-05	Show Text-to-image (T2I) diffusion models such as SDXL and FLUX have achieved impressive photorealism, yet small-scale distortions remain pervasive in limbs, face, text and so on. Existing refinement approaches either perform costly iterative re-generation or rely on vision-language models (VLMs) with weak spatial grounding, leading to semantic drift and unreliable local edits. To close this gap, we propose Agentic Retoucher, a hierarchical decision-driven framework that reformulates post-generation correction as a human-like perception-reasoning-action loop. Specifically, we design (1) a perception agent that learns contextual saliency for fine-grained distortion localization under text-image consistency cues, (2) a reasoning agent that performs human-aligned inferential diagnosis via progressive preference alignment, and (3) an action agent that adaptively plans localized inpainting guided by user preference. This design integrates perceptual evidence, linguistic reasoning, and controllable correction into a unified, self-corrective decision process. To enable fine-grained supervision and quantitative evaluation, we further construct GenBlemish-27K, a dataset of 6K T2I images with 27K annotated artifact regions across 12 categories. Extensive experiments demonstrate that Agentic Retoucher consistently outperforms state-of-the-art methods in perceptual quality, distortion localization and human preference alignment, establishing a new paradigm for self-corrective and perceptually reliable T2I generation.
Thinking with Blueprints: Assisting Vision-Language Models in Spatial Reasoning via Structured Object Representation	2026-01-05	Show Spatial reasoning -- the ability to perceive and reason about relationships in space -- advances vision-language models (VLMs) from visual perception toward spatial semantic understanding. Existing approaches either revisit local image patches, improving fine-grained perception but weakening global spatial awareness, or mark isolated coordinates, which capture object locations but overlook their overall organization. In this work, we integrate the cognitive concept of an object-centric blueprint into VLMs to enhance spatial reasoning. Given an image and a question, the model first constructs a JSON-style blueprint that records the positions, sizes, and attributes of relevant objects, and then reasons over this structured representation to produce the final answer. To achieve this, we introduce three key techniques: (1) blueprint-embedded reasoning traces for supervised fine-tuning to elicit basic reasoning skills; (2) blueprint-aware rewards in reinforcement learning to encourage the blueprint to include an appropriate number of objects and to align final answers with this causal reasoning; and (3) anti-shortcut data augmentation that applies targeted perturbations to images and questions, discouraging reliance on superficial visual or linguistic cues. Experiments show that our method consistently outperforms existing VLMs and specialized spatial reasoning models.	Prepr... Preprint. Under review
Autoregressive Semantic Visual Reconstruction Helps VLMs Understand Better	2026-01-05	Show Typical large vision-language models (LVLMs) apply autoregressive supervision solely to textual sequences, without fully incorporating the visual modality into the learning process. This results in three key limitations: (1) an inability to utilize images without accompanying captions, (2) the risk that captions omit critical visual details, and (3) the challenge that certain vision-centric content cannot be adequately conveyed through text. As a result, current LVLMs often prioritize vision-to-language alignment while potentially overlooking fine-grained visual information. While some prior works have explored autoregressive image generation, effectively leveraging autoregressive visual supervision to enhance image understanding remains an open challenge. In this paper, we introduce Autoregressive Semantic Visual Reconstruction (ASVR), which enables joint learning of visual and textual modalities within a unified autoregressive framework. We show that autoregressively reconstructing the raw visual appearance of images does not enhance and may even impair multimodal understanding. In contrast, autoregressively reconstructing the semantic representation of images consistently improves comprehension. Notably, we find that even when models are given continuous image features as input, they can effectively reconstruct discrete semantic tokens, resulting in stable and consistent improvements across a wide range of multimodal understanding benchmarks. Our approach delivers significant performance gains across varying data scales (556k-2M) and types of LLM bacbones. Specifically, ASVR improves LLaVA-1.5 by 5% in average scores across 14 multimodal benchmarks. The code is available at https://github.com/AlenjandroWang/ASVR.
AFTER: Mitigating the Object Hallucination of LVLM via Adaptive Factual-Guided Activation Editing	2026-01-05	Show Large Vision-Language Models (LVLMs) have achieved substantial progress in cross-modal tasks. However, due to language bias, LVLMs are susceptible to object hallucination, which can be primarily divided into category, attribute, and relation hallucination, significantly impeding the trustworthy AI applications. Editing the internal activations of LVLMs has shown promising effectiveness in mitigating hallucinations with minimal cost. However, previous editing approaches neglect the effective guidance offered by factual textual semantics, thereby struggling to explicitly mitigate language bias. To address these issues, we propose Adaptive Factual-guided Visual-Textual Editing for hallucination mitigation (AFTER), which comprises Factual-Augmented Activation Steering (FAS) and Query-Adaptive Offset Optimization (QAO), to adaptively guides the original biased activations towards factual semantics. Specifically, FAS is proposed to provide factual and general guidance for activation editing, thereby explicitly modeling the precise visual-textual associations. Subsequently, QAO introduces a query-aware offset estimator to establish query-specific editing from the general steering vector, enhancing the diversity and granularity of editing. Extensive experiments on standard hallucination benchmarks across three widely adopted LVLMs validate the efficacy of the proposed AFTER, notably achieving up to a 16.3% reduction of hallucination over baseline on the AMBER benchmark. Our code and data will be released for reproducibility.
Learning Diffusion Policy from Primitive Skills for Robot Manipulation	2026-01-05	Show Diffusion policies (DP) have recently shown great promise for generating actions in robotic manipulation. However, existing approaches often rely on global instructions to produce short-term control signals, which can result in misalignment in action generation. We conjecture that the primitive skills, referred to as fine-grained, short-horizon manipulations, such as move up'' and open the gripper'', provide a more intuitive and effective interface for robot learning. To bridge this gap, we propose SDP, a skill-conditioned DP that integrates interpretable skill learning with conditional action planning. SDP abstracts eight reusable primitive skills across tasks and employs a vision-language model to extract discrete representations from visual observations and language instructions. Based on them, a lightweight router network is designed to assign a desired primitive skill for each state, which helps construct a single-skill policy to generate skill-aligned actions. By decomposing complex tasks into a sequence of primitive skills and selecting a single-skill policy, SDP ensures skill-consistent behavior across diverse tasks. Extensive experiments on two challenging simulation benchmarks and real-world robot deployments demonstrate that SDP consistently outperforms SOTA methods, providing a new paradigm for skill-based robot learning with diffusion policies.	Accepted to AAAI2026
WebGym: Scaling Training Environments for Visual Web Agents with Realistic Tasks	2026-01-05	Show We present WebGym, the largest-to-date open-source environment for training realistic visual web agents. Real websites are non-stationary and diverse, making artificial or small-scale task sets insufficient for robust policy learning. WebGym contains nearly 300,000 tasks with rubric-based evaluations across diverse, real-world websites and difficulty levels. We train agents with a simple reinforcement learning (RL) recipe, which trains on the agent's own interaction traces (rollouts), using task rewards as feedback to guide learning. To enable scaling RL, we speed up sampling of trajectories in WebGym by developing a high-throughput asynchronous rollout system, designed specifically for web agents. Our system achieves a 4-5x rollout speedup compared to naive implementations. Second, we scale the task set breadth, depth, and size, which results in continued performance improvement. Fine-tuning a strong base vision-language model, Qwen-3-VL-8B-Instruct, on WebGym results in an improvement in success rate on an out-of-distribution test set from 26.2% to 42.9%, significantly outperforming agents based on proprietary models such as GPT-4o and GPT-5-Thinking that achieve 27.1% and 29.8%, respectively. This improvement is substantial because our test set consists only of tasks on websites never seen during training, unlike many other prior works on training visual web agents.
MMP-A*: Multimodal Perception Enhanced Incremental Heuristic Search on Path Planning	2026-01-05	Show Autonomous path planning requires a synergy between global reasoning and geometric precision, especially in complex or cluttered environments. While classical A* is valued for its optimality, it incurs prohibitive computational and memory costs in large-scale scenarios. Recent attempts to mitigate these limitations by using Large Language Models for waypoint guidance remain insufficient, as they rely only on text-based reasoning without spatial grounding. As a result, such models often produce incorrect waypoints in topologically complex environments with dead ends, and lack the perceptual capacity to interpret ambiguous physical boundaries. These inconsistencies lead to costly corrective expansions and undermine the intended computational efficiency. We introduce MMP-A*, a multimodal framework that integrates the spatial grounding capabilities of vision-language models with a novel adaptive decay mechanism. By anchoring high-level reasoning in physical geometry, the framework produces coherent waypoint guidance that addresses the limitations of text-only planners. The adaptive decay mechanism dynamically regulates the influence of uncertain waypoints within the heuristic, ensuring geometric validity while substantially reducing memory overhead. To evaluate robustness, we test the framework in challenging environments characterized by severe clutter and topological complexity. Experimental results show that MMP-A* achieves near-optimal trajectories with significantly reduced operational costs, demonstrating its potential as a perception-grounded and computationally efficient paradigm for autonomous navigation.
A Hybrid Architecture for Multi-Stage Claim Document Understanding: Combining Vision-Language Models and Machine Learning for Real-Time Processing	2026-01-05	Show Claims documents are fundamental to healthcare and insurance operations, serving as the basis for reimbursement, auditing, and compliance. However, these documents are typically not born digital; they often exist as scanned PDFs or photographs captured under uncontrolled conditions. Consequently, they exhibit significant content heterogeneity, ranging from typed invoices to handwritten medical reports, as well as linguistic diversity. This challenge is exemplified by operations at Fullerton Health, which handles tens of millions of claims annually across nine markets, including Singapore, the Philippines, Indonesia, Malaysia, Mainland China, Hong Kong, Vietnam, Papua New Guinea, and Cambodia. Such variability, coupled with inconsistent image quality and diverse layouts, poses a significant obstacle to automated parsing and structured information extraction. This paper presents a robust multi-stage pipeline that integrates the multilingual optical character recognition (OCR) engine PaddleOCR, a traditional Logistic Regression classifier, and a compact Vision-Language Model (VLM), Qwen 2.5-VL-7B, to achieve efficient and accurate field extraction from large-scale claims data. The proposed system achieves a document-type classification accuracy of over 95 percent and a field-level extraction accuracy of approximately 87 percent, while maintaining an average processing latency of under 2 seconds per document. Compared to manual processing, which typically requires around 10 minutes per claim, our system delivers a 300x improvement in efficiency. These results demonstrate that combining traditional machine learning models with modern VLMs enables production-grade accuracy and speed for real-world automation. The solution has been successfully deployed in our mobile application and is currently processing tens of thousands of claims weekly from Vietnam and Singapore.	19 pa... 19 pages, 3 figures, 3 tables
Toward Auditable Neuro-Symbolic Reasoning in Pathology: SQL as an Explicit Trace of Evidence	2026-01-05	Show Automated pathology image analysis is central to clinical diagnosis, but clinicians still ask which slide features drive a model's decision and why. Vision-language models can produce natural language explanations, but these are often correlational and lack verifiable evidence. In this paper, we introduce an SQL-centered agentic framework that enables both feature measurement and reasoning to be auditable. Specifically, after extracting human-interpretable cellular features, Feature Reasoning Agents compose and execute SQL queries over feature tables to aggregate visual evidence into quantitative findings. A Knowledge Comparison Agent then evaluates these findings against established pathological knowledge, mirroring how pathologists justify diagnoses from measurable observations. Extensive experiments evaluated on two pathology visual question answering datasets demonstrate our method improves interpretability and decision traceability while producing executable SQL traces that link cellular measurements to diagnostic conclusions.
Entity-Guided Multi-Task Learning for Infrared and Visible Image Fusion	2026-01-05	Show Existing text-driven infrared and visible image fusion approaches often rely on textual information at the sentence level, which can lead to semantic noise from redundant text and fail to fully exploit the deeper semantic value of textual information. To address these issues, we propose a novel fusion approach named Entity-Guided Multi-Task learning for infrared and visible image fusion (EGMT). Our approach includes three key innovative components: (i) A principled method is proposed to extract entity-level textual information from image captions generated by large vision-language models, eliminating semantic noise from raw text while preserving critical semantic information; (ii) A parallel multi-task learning architecture is constructed, which integrates image fusion with a multi-label classification task. By using entities as pseudo-labels, the multi-label classification task provides semantic supervision, enabling the model to achieve a deeper understanding of image content and significantly improving the quality and semantic density of the fused image; (iii) An entity-guided cross-modal interactive module is also developed to facilitate the fine-grained interaction between visual and entity-level textual features, which enhances feature representation by capturing cross-modal dependencies at both inter-visual and visual-entity levels. To promote the wide application of the entity-guided image fusion framework, we release the entity-annotated version of four public datasets (i.e., TNO, RoadScene, M3FD, and MSRS). Extensive experiments demonstrate that EGMT achieves superior performance in preserving salient targets, texture details, and semantic consistency, compared to the state-of-the-art methods. The code and dataset will be publicly available at https://github.com/wyshao-01/EGMT.	Accep... Accepted by IEEE Transactions on Multimedia
SJTU:Spatial judgments in multimodal models towards unified segmentation through coordinate detection	2026-01-05	Show Despite significant advances in vision-language understanding, implementing image segmentation within multimodal architectures remains a fundamental challenge in modern artificial intelligence systems. Existing vision-language models, which primarily rely on backbone architectures or CLIP-based embedding learning, demonstrate inherent limitations in fine-grained spatial localization and operational capabilities. This paper introduces SJTU: Spatial Judgments in Multimodal Models - Towards Unified Segmentation through Coordinate Detection, a framework that leverages spatial coordinate understanding to bridge vision-language interaction and precise segmentation, enabling accurate target identification through natural language instructions. The framework presents an approach for integrating segmentation techniques with vision-language models through spatial inference in multimodal space. By utilizing normalized coordinate detection for bounding boxes and transforming them into actionable segmentation outputs, we establish a connection between spatial and language representations in multimodal architectures. Experimental results demonstrate superior performance across benchmark datasets, achieving IoU scores of 0.5958 on COCO 2017 and 0.6758 on Pascal VOC. Testing on a single NVIDIA RTX 3090 GPU with 512x512 resolution images yields an average inference time of 7 seconds per image, demonstrating the framework's effectiveness in both accuracy and practical deployability. The project code is available at https://github.com/jw-chae/SJTU	A fla... A flaw was discovered in the experimental setup. Therefore, we are retracting the paper
VerLM: Explaining Face Verification Using Natural Language	2026-01-05	Show Face verification systems have seen substantial advancements; however, they often lack transparency in their decision-making processes. In this paper, we introduce an innovative Vision-Language Model (VLM) for Face Verification, which not only accurately determines if two face images depict the same individual but also explicitly explains the rationale behind its decisions. Our model is uniquely trained using two complementary explanation styles: (1) concise explanations that summarize the key factors influencing its decision, and (2) comprehensive explanations detailing the specific differences observed between the images. We adapt and enhance a state-of-the-art modeling approach originally designed for audio-based differentiation to suit visual inputs effectively. This cross-modal transfer significantly improves our model's accuracy and interpretability. The proposed VLM integrates sophisticated feature extraction techniques with advanced reasoning capabilities, enabling clear articulation of its verification process. Our approach demonstrates superior performance, surpassing baseline methods and existing models. These findings highlight the immense potential of vision language models in face verification set up, contributing to more transparent, reliable, and explainable face verification systems.
AdaptInfer: Adaptive Token Pruning for Vision-Language Model Inference with Dynamical Text Guidance	2026-01-05	Show Vision-language models (VLMs) have achieved impressive performance on multimodal reasoning tasks such as visual question answering, image captioning and so on, but their inference cost remains a significant challenge due to the large number of vision tokens processed during the prefill stage. Existing pruning methods often rely on directly using the attention patterns or static text prompt guidance, failing to exploit the dynamic internal signals generated during inference. To address these issues, we propose AdaptInfer, a plug-and-play framework for adaptive vision token pruning in VLMs. First, we introduce a fine-grained, dynamic text-guided pruning mechanism that reuses layer-wise text-to-text attention maps to construct soft priors over text-token importance, allowing more informed scoring of vision tokens at each stage. Second, we perform an offline analysis of cross-modal attention shifts and identify consistent inflection locations in inference, which inspire us to propose a more principled and efficient pruning schedule. Our method is lightweight and plug-and-play, also generalizable across multi-modal tasks. Experimental results have verified the effectiveness of the proposed method. For example, it reduces CUDA latency by 61.3% while maintaining an average accuracy of 93.1% on vanilla LLaVA-1.5-7B. Under the same token budget, AdaptInfer surpasses SOTA in accuracy.
Fine-Grained Preference Optimization Improves Spatial Reasoning in VLMs	2026-01-05	Show Current Vision-Language Models (VLMs) struggle with fine-grained spatial reasoning, particularly when multi-step logic and precise spatial alignment are required. In this work, we introduce SpatialReasoner-R1, a vision-language reasoning model designed to address these limitations. To construct high-quality supervision for spatial reasoning, we design a Multi-Model Monte Carlo Tree Search (M3CTS) method that generates diverse, logically consistent Long Chain-of-Thought (LongCOT) reasoning trajectories. In addition, we propose a fine-grained Direct Preference Optimization (fDPO) method that introduces segment-specific preference granularity for descriptive grounding and logical reasoning, guided by a spatial reward mechanism that evaluates candidate responses based on visual consistency, spatial grounding, and logical coherence. Experimental results demonstrate that fDPO achieves relative performance gains of 4.1% and 9.0% over standard DPO on spatial qualitative and quantitative tasks, respectively. SpatialReasoner-R1, trained with fDPO, sets a new SoTA on SpatialRGPT-Bench, outperforming the strongest baseline by 9.4% in average accuracy, while maintaining competitive performance on general vision-language tasks.
Crafting Adversarial Inputs for Large Vision-Language Models Using Black-Box Optimization	2026-01-05	Show Recent advancements in Large Vision-Language Models (LVLMs) have shown groundbreaking capabilities across diverse multimodal tasks. However, these models remain vulnerable to adversarial jailbreak attacks, where adversaries craft subtle perturbations to bypass safety mechanisms and trigger harmful outputs. Existing white-box attacks methods require full model accessibility, suffer from computing costs and exhibit insufficient adversarial transferability, making them impractical for real-world, black-box settings. To address these limitations, we propose a black-box jailbreak attack on LVLMs via Zeroth-Order optimization using Simultaneous Perturbation Stochastic Approximation (ZO-SPSA). ZO-SPSA provides three key advantages: (i) gradient-free approximation by input-output interactions without requiring model knowledge, (ii) model-agnostic optimization without the surrogate model and (iii) lower resource requirements with reduced GPU memory consumption. We evaluate ZO-SPSA on three LVLMs, including InstructBLIP, LLaVA and MiniGPT-4, achieving the highest jailbreak success rate of 83.0% on InstructBLIP, while maintaining imperceptible perturbations comparable to white-box methods. Moreover, adversarial examples generated from MiniGPT-4 exhibit strong transferability to other LVLMs, with ASR reaching 64.18%. These findings underscore the real-world feasibility of black-box jailbreaks and expose critical weaknesses in the safety mechanisms of current LVLMs	EACL
VisualActBench: Can VLMs See and Act like a Human?	2026-01-04	Show Vision-Language Models (VLMs) have achieved impressive progress in perceiving and describing visual environments. However, their ability to proactively reason and act based solely on visual inputs, without explicit textual prompts, remains underexplored. We introduce a new task, Visual Action Reasoning, and propose VisualActBench, a large-scale benchmark comprising 1,074 videos and 3,733 human-annotated actions across four real-world scenarios. Each action is labeled with an Action Prioritization Level (APL) and a proactive-reactive type to assess models' human-aligned reasoning and value sensitivity. We evaluate 29 VLMs on VisualActBench and find that while frontier models like GPT4o demonstrate relatively strong performance, a significant gap remains compared to human-level reasoning, particularly in generating proactive, high-priority actions. Our results highlight limitations in current VLMs' ability to interpret complex context, anticipate outcomes, and align with human decision-making frameworks. VisualActBench establishes a comprehensive foundation for assessing and improving the real-world readiness of proactive, vision-centric AI agents.
FastV-RAG: Towards Fast and Fine-Grained Video QA with Retrieval-Augmented Generation	2026-01-04	Show Vision-Language Models (VLMs) excel at visual reasoning but still struggle with integrating external knowledge. Retrieval-Augmented Generation (RAG) is a promising solution, but current methods remain inefficient and often fail to maintain high answer quality. To address these challenges, we propose VideoSpeculateRAG, an efficient VLM-based RAG framework built on two key ideas. First, we introduce a speculative decoding pipeline: a lightweight draft model quickly generates multiple answer candidates, which are then verified and refined by a more accurate heavyweight model, substantially reducing inference latency without sacrificing correctness. Second, we identify a major source of error - incorrect entity recognition in retrieved knowledge - and mitigate it with a simple yet effective similarity-based filtering strategy that improves entity alignment and boosts overall answer accuracy. Experiments demonstrate that VideoSpeculateRAG achieves comparable or higher accuracy than standard RAG approaches while accelerating inference by approximately 2x. Our framework highlights the potential of combining speculative decoding with retrieval-augmented reasoning to enhance efficiency and reliability in complex, knowledge-intensive multimodal tasks.
Unified Generation and Self-Verification for Vision-Language Models via Advantage Decoupled Preference Optimization	2026-01-04	Show Parallel test-time scaling typically trains separate generation and verification models, incurring high training and inference costs. We propose Advantage Decoupled Preference Optimization (ADPO), a unified reinforcement learning framework that jointly learns answer generation and self-verification within a single policy. ADPO introduces two innovations: a preference verification reward improving verification capability and a decoupled optimization mechanism enabling synergistic optimization of generation and verification. Specifically, the preference verification reward computes mean verification scores from positive and negative samples as decision thresholds, providing positive feedback when prediction correctness aligns with answer correctness. Meanwhile, the advantage decoupled optimization computes separate advantages for generation and verification, applies token masks to isolate gradients, and combines masked GRPO objectives, preserving generation quality while calibrating verification scores. ADPO achieves up to +34.1% higher verification AUC and -53.5% lower inference time, with significant gains of +2.8%/+1.4% accuracy on MathVista/MMMU, +1.9 cIoU on ReasonSeg, and +1.7%/+1.0% step success rate on AndroidControl/GUI Odyssey.
How to make Medical AI Systems safer? Simulating Vulnerabilities, and Threats in Multimodal Medical RAG System	2026-01-04	Show Large Vision-Language Models (LVLMs) augmented with Retrieval-Augmented Generation (RAG) are increasingly employed in medical AI to enhance factual grounding through external clinical image-text retrieval. However, this reliance creates a significant attack surface. We propose MedThreatRAG, a novel multimodal poisoning framework that systematically probes vulnerabilities in medical RAG systems by injecting adversarial image-text pairs. A key innovation of our approach is the construction of a simulated semi-open attack environment, mimicking real-world medical systems that permit periodic knowledge base updates via user or pipeline contributions. Within this setting, we introduce and emphasize Cross-Modal Conflict Injection (CMCI), which embeds subtle semantic contradictions between medical images and their paired reports. These mismatches degrade retrieval and generation by disrupting cross-modal alignment while remaining sufficiently plausible to evade conventional filters. While basic textual and visual attacks are included for completeness, CMCI demonstrates the most severe degradation. Evaluations on IU-Xray and MIMIC-CXR QA tasks show that MedThreatRAG reduces answer F1 scores by up to 27.66% and lowers LLaVA-Med-1.5 F1 rates to as low as 51.36%. Our findings expose fundamental security gaps in clinical RAG systems and highlight the urgent need for threat-aware design and robust multimodal consistency checks. Finally, we conclude with a concise set of guidelines to inform the safe development of future multimodal medical RAG systems.	Sumbi... Sumbitted to 2026 ICASSP
AirSpatialBot: A Spatially-Aware Aerial Agent for Fine-Grained Vehicle Attribute Recognization and Retrieval	2026-01-04	Show Despite notable advancements in remote sensing vision-language models (VLMs), existing models often struggle with spatial understanding, limiting their effectiveness in real-world applications. To push the boundaries of VLMs in remote sensing, we specifically address vehicle imagery captured by drones and introduce a spatially-aware dataset AirSpatial, which comprises over 206K instructions and introduces two novel tasks: Spatial Grounding and Spatial Question Answering. It is also the first remote sensing grounding dataset to provide 3DBB. To effectively leverage existing image understanding of VLMs to spatial domains, we adopt a two-stage training strategy comprising Image Understanding Pre-training and Spatial Understanding Fine-tuning. Utilizing this trained spatially-aware VLM, we develop an aerial agent, AirSpatialBot, which is capable of fine-grained vehicle attribute recognition and retrieval. By dynamically integrating task planning, image understanding, spatial understanding, and task execution capabilities, AirSpatialBot adapts to diverse query requirements. Experimental results validate the effectiveness of our approach, revealing the spatial limitations of existing VLMs while providing valuable insights. The model, code, and datasets will be released at https://github.com/VisionXLab/AirSpatialBot	12 pages, 9 figures
Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbone	2026-01-04	Show While autoregressive Large Vision-Language Models (VLMs) have achieved remarkable success, their sequential generation often limits their efficacy in complex visual planning and dynamic robotic control. In this work, we investigate the potential of constructing Vision-Language Models upon diffusion-based large language models (dLLMs) to overcome these limitations. We introduce Dream-VL, an open diffusion-based VLM (dVLM) that achieves state-of-the-art performance among previous dVLMs. Dream-VL is comparable to top-tier AR-based VLMs trained on open data on various benchmarks but exhibits superior potential when applied to visual planning tasks. Building upon Dream-VL, we introduce Dream-VLA, a dLLM-based Vision-Language-Action model (dVLA) developed through continuous pre-training on open robotic datasets. We demonstrate that the natively bidirectional nature of this diffusion backbone serves as a superior foundation for VLA tasks, inherently suited for action chunking and parallel generation, leading to significantly faster convergence in downstream fine-tuning. Dream-VLA achieves top-tier performance of 97.2% average success rate on LIBERO, 71.4% overall average on SimplerEnv-Bridge, and 60.5% overall average on SimplerEnv-Fractal, surpassing leading models such as $π_0$ and GR00T-N1. We also validate that dVLMs surpass AR baselines on downstream tasks across different training objectives. We release both Dream-VL and Dream-VLA to facilitate further research in the community.	Add r... Add real-world experiments
Multimodal Fact-Checking: An Agent-based Approach	2026-01-04	Show The rapid spread of multimodal misinformation poses a growing challenge for automated fact-checking systems. Existing approaches, including large vision language models (LVLMs) and deep multimodal fusion methods, often fall short due to limited reasoning and shallow evidence utilization. A key bottleneck is the lack of dedicated datasets that provide complete real-world multimodal misinformation instances accompanied by annotated reasoning processes and verifiable evidence. To address this limitation, we introduce RW-Post, a high-quality and explainable dataset for real-world multimodal fact-checking. RW-Post aligns real-world multimodal claims with their original social media posts, preserving the rich contextual information in which the claims are made. In addition, the dataset includes detailed reasoning and explicitly linked evidence, which are derived from human written fact-checking articles via a large language model assisted extraction pipeline, enabling comprehensive verification and explanation. Building upon RW-Post, we propose AgentFact, an agent-based multimodal fact-checking framework designed to emulate the human verification workflow. AgentFact consists of five specialized agents that collaboratively handle key fact-checking subtasks, including strategy planning, high-quality evidence retrieval, visual analysis, reasoning, and explanation generation. These agents are orchestrated through an iterative workflow that alternates between evidence searching and task-aware evidence filtering and reasoning, facilitating strategic decision-making and systematic evidence analysis. Extensive experimental results demonstrate that the synergy between RW-Post and AgentFact substantially improves both the accuracy and interpretability of multimodal fact-checking.	Code ... Code and dataset will be released at https://github.com/xudanni0927/AgentFact
TraveLLaMA: A Multimodal Travel Assistant with Large-Scale Dataset and Structured Reasoning	2026-01-04	Show Tourism and travel planning increasingly rely on digital assistance, yet existing multimodal AI systems often lack specialized knowledge and contextual understanding of urban environments. We present TraveLLaMA, a specialized multimodal language model designed for comprehensive travel assistance. Our work addresses the fundamental challenge of developing practical AI travel assistants through three key contributions: (1) TravelQA, a novel dataset of 265k question-answer pairs combining 160k text QA from authentic travel sources, 100k vision-language QA featuring maps and location imagery, and 5k expert-annotated Chain-of-Thought reasoning examples; (2) Travel-CoT, a structured reasoning framework that decomposes travel queries into spatial, temporal, and practical dimensions, improving answer accuracy by 10.8% while providing interpretable decision paths; and (3) an interactive agent system validated through extensive user studies. Through fine-tuning experiments on state-of-the-art vision-language models (LLaVA, Qwen-VL, Shikra), we achieve 6.2-9.4% base improvements, further enhanced by Travel-CoT reasoning. Our model demonstrates superior capabilities in contextual travel recommendations, map interpretation, and scene understanding while providing practical information such as operating hours and cultural insights. User studies with 500 participants show TraveLLaMA achieves a System Usability Scale score of 82.5, significantly outperforming general-purpose models and establishing new standards for multimodal travel assistance systems.	AAAI 2026 Oral
LSRE: Latent Semantic Rule Encoding for Real-Time Semantic Risk Detection in Autonomous Driving	2026-01-04	Show Real-world autonomous driving must adhere to complex human social rules that extend beyond legally codified traffic regulations. Many of these semantic constraints, such as yielding to emergency vehicles, complying with traffic officers' gestures, or stopping for school buses, are intuitive for humans yet difficult to encode explicitly. Although large vision-language models (VLMs) can interpret such semantics, their inference cost makes them impractical for real-time deployment. This work proposes LSRE, a Latent Semantic Rule Encoding framework that converts sparsely sampled VLM judgments into decision boundaries within the latent space of a recurrent world model. By encoding language-defined safety semantics into a lightweight latent classifier, LSRE enables real-time semantic risk assessment at 10 Hz without per-frame VLM queries. Experiments on six semantic-failure scenarios in CARLA demonstrate that LSRE attains semantic risk detection accuracy comparable to a large VLM baseline, while providing substantially earlier hazard anticipation and maintaining low computational latency. LSRE further generalizes to rarely seen semantic-similar test cases, indicating that language-guided latent classification offers an effective and deployable mechanism for semantic safety monitoring in autonomous driving.
LinMU: Multimodal Understanding Made Linear	2026-01-04	Show Modern Vision-Language Models (VLMs) achieve impressive performance but are limited by the quadratic complexity of self-attention, which prevents their deployment on edge devices and makes their understanding of high-resolution images and long-context videos prohibitively expensive. To address this challenge, we introduce LinMU (Linear-complexity Multimodal Understanding), a VLM design that achieves linear complexity without using any quadratic-complexity modules while maintaining the performance of global-attention-based VLMs. LinMU replaces every self-attention layer in the VLM with the M-MATE block: a dual-branch module that combines a bidirectional state-space model for global context (Flex-MA branch) with localized Swin-style window attention (Local-Swin branch) for adjacent correlations. To transform a pre-trained VLM into the LinMU architecture, we propose a three-stage distillation framework that (i) initializes both branches with self-attention weights and trains the Flex-MA branch alone, (ii) unfreezes the Local-Swin branch and fine-tunes it jointly with the Flex-MA branch, and (iii) unfreezes the remaining blocks and fine-tunes them using LoRA adapters, while regressing on hidden states and token-level logits of the frozen VLM teacher. On MMMU, TextVQA, LongVideoBench, Video-MME, and other benchmarks, LinMU matches the performance of teacher models, yet reduces Time-To-First-Token (TTFT) by up to 2.7$\times$ and improves token throughput by up to 9.0$\times$ on minute-length videos. Ablations confirm the importance of each distillation stage and the necessity of the two branches of the M-MATE block. The proposed framework demonstrates that state-of-the-art multimodal reasoning can be achieved without quadratic attention, thus opening up avenues for long-context VLMs that can deal with high-resolution images and long videos.	23 pages, 7 figures
RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics	2026-01-03	Show Spatial referring is a fundamental capability of embodied robots to interact with the 3D physical world. However, even with the powerful pretrained vision language models (VLMs), recent approaches are still not qualified to accurately understand the complex 3D scenes and dynamically reason about the instruction-indicated locations for interaction. To this end, we propose RoboRefer, a 3D-aware VLM that can first achieve precise spatial understanding by integrating a disentangled but dedicated depth encoder via supervised fine-tuning (SFT). Moreover, RoboRefer advances generalized multi-step spatial reasoning via reinforcement fine-tuning (RFT), with metric-sensitive process reward functions tailored for spatial referring tasks. To support SFT and RFT training, we introduce RefSpatial, a large-scale dataset of 20M QA pairs (2x prior), covering 31 spatial relations (vs. 15 prior) and supporting complex reasoning processes (up to 5 steps). In addition, we introduce RefSpatial-Bench, a challenging benchmark filling the gap in evaluating spatial referring with multi-step reasoning. Experiments show that SFT-trained RoboRefer achieves state-of-the-art spatial understanding, with an average success rate of 89.6%. RFT-trained RoboRefer further outperforms all other baselines by a large margin, even surpassing Gemini-2.5-Pro by 17.4% in average accuracy on RefSpatial-Bench. Notably, RoboRefer can be integrated with various control policies to execute long-horizon, dynamic tasks across diverse robots (e,g., UR5, G1 humanoid) in cluttered real-world scenes.	Accep... Accepted by NeurIPS 2025. Project page: https://zhoues.github.io/RoboRefer/
Performance Gap in Entity Knowledge Extraction Across Modalities in Vision Language Models	2026-01-03	Show Vision-language models (VLMs) excel at extracting and reasoning about information from images. Yet, their capacity to leverage internal knowledge about specific entities remains underexplored. This work investigates the disparity in model performance when answering factual questions about an entity described in text versus depicted in an image. Our results reveal a significant accuracy drop - reaching 18% for some models - when the entity is presented visually instead of textually. To study this gap we present PopVQA, a dataset which allows separating entity recognition and question answering, and use it to benchmark several models. We hypothesize that this decline arises from limitations in how information flows from image tokens to query tokens. Thus, we use mechanistic interpretability tools to reveal that, although image tokens are preprocessed by the vision encoder, meaningful information flow from these tokens occurs only in the much deeper layers. Furthermore, critical image processing happens in the language model's middle layers, allowing few layers for consecutive reasoning, highlighting a potential inefficiency in how the model utilizes its layers for reasoning. These insights shed light on the internal mechanics of VLMs and offer pathways for enhancing their reasoning capabilities. PopVQA can be found at https://huggingface.co/datasets/idoco/PopVQA.	Accep... Accepted to ACL 2025 Main Conference
DISCODE: Distribution-Aware Score Decoder for Robust Automatic Evaluation of Image Captioning	2026-01-03	Show Large vision-language models (LVLMs) have shown impressive performance across a broad range of multimodal tasks. However, robust image caption evaluation using LVLMs remains challenging, particularly under domain-shift scenarios. To address this issue, we introduce the Distribution-Aware Score Decoder (DISCODE), a novel finetuning-free method that generates robust evaluation scores better aligned with human judgments across diverse domains. The core idea behind DISCODE lies in its test-time adaptive evaluation approach, which introduces the Adaptive Test-Time (ATT) loss, leveraging a Gaussian prior distribution to improve robustness in evaluation score estimation. This loss is efficiently minimized at test time using an analytical solution that we derive. Furthermore, we introduce the Multi-domain Caption Evaluation (MCEval) benchmark, a new image captioning evaluation benchmark covering six distinct domains, designed to assess the robustness of evaluation metrics. In our experiments, we demonstrate that DISCODE achieves state-of-the-art performance as a reference-free evaluation metric across MCEval and four representative existing benchmarks.	Paper... Paper accepted to AAAI 2026
SPoRC-VIST: A Benchmark for Evaluating Generative Natural Narrative in Vision-Language Models	2026-01-03	Show Vision-Language Models (VLMs) have achieved remarkable success in descriptive tasks such as image captioning and visual question answering (VQA). However, their ability to generate engaging, long-form narratives -- specifically multi-speaker podcast dialogues -- remains under-explored and difficult to evaluate. Standard metrics like BLEU and ROUGE fail to capture the nuances of conversational naturalness, personality, and narrative flow, often rewarding safe, repetitive outputs over engaging storytelling. In this work, we present a novel pipeline for end-to-end visual podcast generation, and fine-tune a Qwen3-VL-32B model on a curated dataset of 4,000 image-dialogue pairs. Crucially, we use a synthetic-to-real training strategy: we train on high-quality podcast dialogues from the Structured Podcast Research Corpus (SPoRC) paired with synthetically generated imagery, and evaluate on real-world photo sequences from the Visual Storytelling Dataset (VIST). This rigorous setup tests the model's ability to generalize from synthetic training data to real-world visual domains. We propose a comprehensive evaluation framework that moves beyond textual overlap, and use AI-as-a-judge (Gemini 3 Pro, Claude Opus 4.5, GPT 5.2) and novel style metrics (average turn length, speaker switch rate) to assess quality. Our experiments demonstrate that our fine-tuned 32B model significantly outperforms a 235B base model in conversational naturalness ($>$80% win rate) and narrative depth (+50% turn length), while maintaining identical visual grounding capabilities (CLIPScore: 20.39).	14 pa... 14 pages, 3 figures. Accepted to WVAQ 2026, WACV 2026

world model

Title	Date	Abstract	Comment
Indicating Robot Vision Capabilities with Augmented Reality	2026-01-06	Show Research indicates that humans can mistakenly assume that robots and humans have the same field of view, possessing an inaccurate mental model of robots. This misperception may lead to failures during human-robot collaboration tasks where robots might be asked to complete impossible tasks about out-of-view objects. The issue is more severe when robots do not have a chance to scan the scene to update their world model while focusing on assigned tasks. To help align humans' mental models of robots' vision capabilities, we propose four field-of-view indicators in augmented reality and conducted a human-subjects experiment (N=41) to evaluate them in a collaborative assembly task regarding accuracy, confidence, task efficiency, and workload. These indicators span a spectrum of positions: two at robot's eye and head space -- deepening eye socket and adding blocks to two sides of the eyes (i.e., egocentric), and two anchoring in the robot's task space -- adding extended blocks from the sides of eyes to the table and placing blocks directly on the tables (i.e., allocentric). Results showed that, when placed directly in the task space, the allocentric indicator yields the highest accuracy, although with a delay in interpreting the robot's field of view. When placed at the robot's eyes, the egocentric indicator of deeper eye sockets, possible for physical alteration, also increased accuracy. In all indicators, participants' confidence was high while cognitive load remained low. Finally, we contribute six guidelines for practitioners to apply our augmented reality indicators or physical alterations to align humans' mental models with robots' vision capabilities.
Evaluating Gemini Robotics Policies in a Veo World Simulator	2026-01-06	Show Generative world models hold significant potential for simulating interactions with visuomotor policies in varied environments. Frontier video models can enable generation of realistic observations and environment interactions in a scalable and general manner. However, the use of video models in robotics has been limited primarily to in-distribution evaluations, i.e., scenarios that are similar to ones used to train the policy or fine-tune the base video model. In this report, we demonstrate that video models can be used for the entire spectrum of policy evaluation use cases in robotics: from assessing nominal performance to out-of-distribution (OOD) generalization, and probing physical and semantic safety. We introduce a generative evaluation system built upon a frontier video foundation model (Veo). The system is optimized to support robot action conditioning and multi-view consistency, while integrating generative image-editing and multi-view completion to synthesize realistic variations of real-world scenes along multiple axes of generalization. We demonstrate that the system preserves the base capabilities of the video model to enable accurate simulation of scenes that have been edited to include novel interaction objects, novel visual backgrounds, and novel distractor objects. This fidelity enables accurately predicting the relative performance of different policies in both nominal and OOD conditions, determining the relative impact of different axes of generalization on policy performance, and performing red teaming of policies to expose behaviors that violate physical or semantic safety constraints. We validate these capabilities through 1600+ real-world evaluations of eight Gemini Robotics policy checkpoints and five tasks for a bimanual manipulator.
Time-Scaling Is What Agents Need Now	2026-01-06	Show Early artificial intelligence paradigms exhibited separated cognitive functions: Neural Networks focused on "perception-representation," Reinforcement Learning on "decision-making-behavior," and Symbolic AI on "knowledge-reasoning." With Transformer-based large models and world models, these paradigms are converging into cognitive agents with closed-loop "perception-decision-action" capabilities. Humans solve complex problems under limited cognitive resources through temporalized sequential reasoning. Language relies on problem space search for deep semantic reasoning. While early large language models (LLMs) could generate fluent text, they lacked robust semantic reasoning capabilities. Prompting techniques like Chain-of-Thought (CoT) and Tree-of-Thought (ToT) extended reasoning paths by making intermediate steps explicit. Recent models like DeepSeek-R1 enhanced performance through explicit reasoning trajectories. However, these methods have limitations in search completeness and efficiency. This highlights the need for "Time-Scaling"--the systematic extension and optimization of an agent's ability to unfold reasoning over time. Time-Scaling refers to architectural design utilizing extended temporal pathways, enabling deeper problem space exploration, dynamic strategy adjustment, and enhanced metacognitive control, paralleling human sequential reasoning under cognitive constraints. It represents a critical frontier for enhancing deep reasoning and problem-solving without proportional increases in static model parameters. Advancing intelligent agent capabilities requires placing Time-Scaling principles at the forefront, positioning explicit temporal reasoning management as foundational.
InternVLA-A1: Unifying Understanding, Generation and Action for Robotic Manipulation	2026-01-05	Show Prevalent Vision-Language-Action (VLA) models are typically built upon Multimodal Large Language Models (MLLMs) and demonstrate exceptional proficiency in semantic understanding, but they inherently lack the capability to deduce physical world dynamics. Consequently, recent approaches have shifted toward World Models, typically formulated via video prediction; however, these methods often suffer from a lack of semantic grounding and exhibit brittleness when handling prediction errors. To synergize semantic understanding with dynamic predictive capabilities, we present InternVLA-A1. This model employs a unified Mixture-of-Transformers architecture, coordinating three experts for scene understanding, visual foresight generation, and action execution. These components interact seamlessly through a unified masked self-attention mechanism. Building upon InternVL3 and Qwen3-VL, we instantiate InternVLA-A1 at 2B and 3B parameter scales. We pre-train these models on hybrid synthetic-real datasets spanning InternData-A1 and Agibot-World, covering over 533M frames. This hybrid training strategy effectively harnesses the diversity of synthetic simulation data while minimizing the sim-to-real gap. We evaluated InternVLA-A1 across 12 real-world robotic tasks and simulation benchmark. It significantly outperforms leading models like pi0 and GR00T N1.5, achieving a 14.5% improvement in daily tasks and a 40%-73.3% boost in dynamic settings, such as conveyor belt sorting.	Homep... Homepage: https://internrobotics.github.io/internvla-a1.github.io/
Grounded Test-Time Adaptation for LLM Agents	2026-01-05	Show Large language model (LLM)-based agents struggle to generalize to novel and complex environments, such as unseen websites or new sets of functions, due to a fundamental mismatch between their pre-training and test-time conditions. This challenge stems from two distinct failure modes: a syntactic misunderstanding of environment-specific components like observation formats, and a semantic misunderstanding of state-transition dynamics, which are only revealed at test time. To address these issues, we propose two distinct and complementary strategies for adapting LLM agents by leveraging environment-specific information available during deployment. First, an online distributional adaptation method parameterizes environmental nuances by learning a lightweight adaptation vector that biases the model's output distribution, enabling rapid alignment with an environment response format. Second, a deployment-time dynamics grounding method employs a persona-driven exploration phase to systematically probe and learn the environment's causal dynamics before task execution, equipping the agent with a nonparametric world model. We evaluate these strategies across diverse agentic benchmarks, including function calling and web navigation. Our empirical results show the effectiveness of both strategies across all benchmarks with minimal computational cost. We find that dynamics grounding is particularly effective in complex environments where unpredictable dynamics pose a major obstacle, demonstrating a robust path toward more generalizable and capable LLM-based agents. For example, on the WebArena multi-site split, this method increases the agent's success rate from 2% to 23%.	Our c... Our code is available here: https://github.com/r2llab/GTTA
SurgWorld: Learning Surgical Robot Policies from Videos via World Modeling	2026-01-05	Show Data scarcity remains a fundamental barrier to achieving fully autonomous surgical robots. While large scale vision language action (VLA) models have shown impressive generalization in household and industrial manipulation by leveraging paired video action data from diverse domains, surgical robotics suffers from the paucity of datasets that include both visual observations and accurate robot kinematics. In contrast, vast corpora of surgical videos exist, but they lack corresponding action labels, preventing direct application of imitation learning or VLA training. In this work, we aim to alleviate this problem by learning policy models from SurgWorld, a world model designed for surgical physical AI. We curated the Surgical Action Text Alignment (SATA) dataset with detailed action description specifically for surgical robots. Then we built SurgeWorld based on the most advanced physical AI world model and SATA. It's able to generate diverse, generalizable and realistic surgery videos. We are also the first to use an inverse dynamics model to infer pseudokinematics from synthetic surgical videos, producing synthetic paired video action data. We demonstrate that a surgical VLA policy trained with these augmented data significantly outperforms models trained only on real demonstrations on a real surgical robot platform. Our approach offers a scalable path toward autonomous surgical skill acquisition by leveraging the abundance of unlabeled surgical video and generative world modeling, thus opening the door to generalizable and data efficient surgical robot policies.
Long-Horizon Model-Based Offline Reinforcement Learning Without Conservatism	2026-01-05	Show Popular offline reinforcement learning (RL) methods rely on conservatism, either by penalizing out-of-dataset actions or by restricting rollout horizons. In this work, we question the universality of this principle and instead revisit a complementary one: a Bayesian perspective. Rather than enforcing conservatism, the Bayesian approach tackles epistemic uncertainty in offline data by modeling a posterior distribution over plausible world models and training a history-dependent agent to maximize expected rewards, enabling test-time generalization. We first illustrate, in a bandit setting, that Bayesianism excels on low-quality datasets where conservatism fails. We then scale this principle to realistic tasks and show that long-horizon planning is critical for reducing value overestimation once conservatism is removed. To make this feasible, we introduce key design choices for performing and learning from long-horizon rollouts while controlling compounding errors. These yield our algorithm, NEUBAY, grounded in the neutral Bayesian principle. On D4RL and NeoRL benchmarks, NEUBAY generally matches or surpasses leading conservative algorithms, achieving new state-of-the-art on 7 datasets. Notably, it succeeds with rollout horizons of several hundred steps, contrary to dominant practice. Finally, we characterize datasets by quality and coverage, showing when NEUBAY is preferable to conservative methods. Together, we argue NEUBAY lays the foundation for a new practical direction in offline and model-based RL.	Prepr... Preprint (52 pages, 15 figures) and code is available at https://github.com/twni2016/neubay
AI Agent Systems: Architectures, Applications, and Evaluation	2026-01-05	Show AI agents -- systems that combine foundation models with reasoning, planning, memory, and tool use -- are rapidly becoming a practical interface between natural-language intent and real-world computation. This survey synthesizes the emerging landscape of AI agent architectures across: (i) deliberation and reasoning (e.g., chain-of-thought-style decomposition, self-reflection and verification, and constraint-aware decision making), (ii) planning and control (from reactive policies to hierarchical and multi-step planners), and (iii) tool calling and environment interaction (retrieval, code execution, APIs, and multimodal perception). We organize prior work into a unified taxonomy spanning agent components (policy/LLM core, memory, world models, planners, tool routers, and critics), orchestration patterns (single-agent vs.\ multi-agent; centralized vs.\ decentralized coordination), and deployment settings (offline analysis vs.\ online interactive assistance; safety-critical vs.\ open-ended tasks). We discuss key design trade-offs -- latency vs.\ accuracy, autonomy vs.\ controllability, and capability vs.\ reliability -- and highlight how evaluation is complicated by non-determinism, long-horizon credit assignment, tool and environment variability, and hidden costs such as retries and context growth. Finally, we summarize measurement and benchmarking practices (task suites, human preference and utility metrics, success under constraints, robustness and security) and identify open challenges including verification and guardrails for tool actions, scalable memory and context management, interpretability of agent decisions, and reproducible evaluation under realistic workloads.
Explicit World Models for Reliable Human-Robot Collaboration	2026-01-05	Show This paper addresses the topic of robustness under sensing noise, ambiguous instructions, and human-robot interaction. We take a radically different tack to the issue of reliable embodied AI: instead of focusing on formal verification methods aimed at achieving model predictability and robustness, we emphasise the dynamic, ambiguous and subjective nature of human-robot interactions that requires embodied AI systems to perceive, interpret, and respond to human intentions in a manner that is consistent, comprehensible and aligned with human expectations. We argue that when embodied agents operate in human environments that are inherently social, multimodal, and fluid, reliability is contextually determined and only has meaning in relation to the goals and expectations of humans involved in the interaction. This calls for a fundamentally different approach to achieving reliable embodied AI that is centred on building and updating an accessible "explicit world model" representing the common ground between human and AI, that is used to align robot behaviours with human expectations.	Accep... Accepted to AAAI-26 Bridge Program B10: Making Embodied AI Reliable with Testing and Formal Verification
Coupled Distributional Random Expert Distillation for World Model Online Imitation Learning	2026-01-04	Show Imitation Learning (IL) has achieved remarkable success across various domains, including robotics, autonomous driving, and healthcare, by enabling agents to learn complex behaviors from expert demonstrations. However, existing IL methods often face instability challenges, particularly when relying on adversarial reward or value formulations in world model frameworks. In this work, we propose a novel approach to online imitation learning that addresses these limitations through a reward model based on random network distillation (RND) for density estimation. Our reward model is built on the joint estimation of expert and behavioral distributions within the latent space of the world model. We evaluate our method across diverse benchmarks, including DMControl, Meta-World, and ManiSkill2, showcasing its ability to deliver stable performance and achieve expert-level results in both locomotion and manipulation tasks. Our approach demonstrates improved stability over adversarial methods while maintaining expert-level performance.	NeurI... NeurIPS 2025 Workshop of Embodied World Models; Code Available at: https://github.com/TobyLeelsz/CDRED-WM
HanoiWorld : A Joint Embedding Predictive Architecture BasedWorld Model for Autonomous Vehicle Controller	2026-01-04	Show Current attempts of Reinforcement Learning for Autonomous Controller are data-demanding while the results are under-performed, unstable, and unable to grasp and anchor on the concept of safety, and over-concentrating on noise features due to the nature of pixel reconstruction. While current Self-Supervised Learningapproachs that learning on high-dimensional representations by leveraging the JointEmbedding Predictive Architecture (JEPA) are interesting and an effective alternative, as the idea mimics the natural ability of the human brain in acquiring new skill usingimagination and minimal samples of observations. This study introduces Hanoi-World, a JEPA-based world model that using recurrent neural network (RNN) formaking longterm horizontal planning with effective inference time. Experimentsconducted on the Highway-Env package with difference enviroment showcase the effective capability of making a driving plan while safety-awareness, with considerablecollision rate in comparison with SOTA baselines
DrivingGen: A Comprehensive Benchmark for Generative Video World Models in Autonomous Driving	2026-01-04	Show Video generation models, as one form of world models, have emerged as one of the most exciting frontiers in AI, promising agents the ability to imagine the future by modeling the temporal evolution of complex scenes. In autonomous driving, this vision gives rise to driving world models: generative simulators that imagine ego and agent futures, enabling scalable simulation, safe testing of corner cases, and rich synthetic data generation. Yet, despite fast-growing research activity, the field lacks a rigorous benchmark to measure progress and guide priorities. Existing evaluations remain limited: generic video metrics overlook safety-critical imaging factors; trajectory plausibility is rarely quantified; temporal and agent-level consistency is neglected; and controllability with respect to ego conditioning is ignored. Moreover, current datasets fail to cover the diversity of conditions required for real-world deployment. To address these gaps, we present DrivingGen, the first comprehensive benchmark for generative driving world models. DrivingGen combines a diverse evaluation dataset curated from both driving datasets and internet-scale video sources, spanning varied weather, time of day, geographic regions, and complex maneuvers, with a suite of new metrics that jointly assess visual realism, trajectory plausibility, temporal coherence, and controllability. Benchmarking 14 state-of-the-art models reveals clear trade-offs: general models look better but break physics, while driving-specific ones capture motion realistically but lag in visual quality. DrivingGen offers a unified evaluation framework to foster reliable, controllable, and deployable driving world models, enabling scalable simulation, planning, and data-driven decision-making.	10 pa... 10 pages, 4 figures; Project Website: https://drivinggen-bench.github.io/
LSRE: Latent Semantic Rule Encoding for Real-Time Semantic Risk Detection in Autonomous Driving	2026-01-04	Show Real-world autonomous driving must adhere to complex human social rules that extend beyond legally codified traffic regulations. Many of these semantic constraints, such as yielding to emergency vehicles, complying with traffic officers' gestures, or stopping for school buses, are intuitive for humans yet difficult to encode explicitly. Although large vision-language models (VLMs) can interpret such semantics, their inference cost makes them impractical for real-time deployment. This work proposes LSRE, a Latent Semantic Rule Encoding framework that converts sparsely sampled VLM judgments into decision boundaries within the latent space of a recurrent world model. By encoding language-defined safety semantics into a lightweight latent classifier, LSRE enables real-time semantic risk assessment at 10 Hz without per-frame VLM queries. Experiments on six semantic-failure scenarios in CARLA demonstrate that LSRE attains semantic risk detection accuracy comparable to a large VLM baseline, while providing substantially earlier hazard anticipation and maintaining low computational latency. LSRE further generalizes to rarely seen semantic-similar test cases, indicating that language-guided latent classification offers an effective and deployable mechanism for semantic safety monitoring in autonomous driving.
Digital Twin AI: Opportunities and Challenges from Large Language Models to World Models	2026-01-04	Show Digital twins, as precise digital representations of physical systems, have evolved from passive simulation tools into intelligent and autonomous entities through the integration of artificial intelligence technologies. This paper presents a unified four-stage framework that systematically characterizes AI integration across the digital twin lifecycle, spanning modeling, mirroring, intervention, and autonomous management. By synthesizing existing technologies and practices, we distill a unified four-stage framework that systematically characterizes how AI methodologies are embedded across the digital twin lifecycle: (1) modeling the physical twin through physics-based and physics-informed AI approaches, (2) mirroring the physical system into a digital twin with real-time synchronization, (3) intervening in the physical twin through predictive modeling, anomaly detection, and optimization strategies, and (4) achieving autonomous management through large language models, foundation models, and intelligent agents. We analyze the synergy between physics-based modeling and data-driven learning, highlighting the shift from traditional numerical solvers to physics-informed and foundation models for physical systems. Furthermore, we examine how generative AI technologies, including large language models and generative world models, transform digital twins into proactive and self-improving cognitive systems capable of reasoning, communication, and creative scenario generation. Through a cross-domain review spanning eleven application domains, including healthcare, aerospace, smart manufacturing, robotics, and smart cities, we identify common challenges related to scalability, explainability, and trustworthiness, and outline directions for responsible AI-driven digital twin systems.
Flow Equivariant World Models: Memory for Partially Observed Dynamic Environments	2026-01-03	Show Embodied systems experience the world as 'a symphony of flows': a combination of many continuous streams of sensory input coupled to self-motion, interwoven with the dynamics of external objects. These streams obey smooth, time-parameterized symmetries, which combine through a precisely structured algebra; yet most neural network world models ignore this structure and instead repeatedly re-learn the same transformations from data. In this work, we introduce 'Flow Equivariant World Models', a framework in which both self-motion and external object motion are unified as one-parameter Lie group 'flows'. We leverage this unification to implement group equivariance with respect to these transformations, thereby providing a stable latent world representation over hundreds of timesteps. On both 2D and 3D partially observed video world modeling benchmarks, we demonstrate that Flow Equivariant World Models significantly outperform comparable state-of-the-art diffusion-based and memory-augmented world modeling architectures -- particularly when there are predictable world dynamics outside the agent's current field of view. We show that flow equivariance is particularly beneficial for long rollouts, generalizing far beyond the training horizon. By structuring world model representations with respect to internal and external motion, flow equivariance charts a scalable route to data efficient, symmetry-guided, embodied intelligence. Project link: https://flowequivariantworldmodels.github.io.	11 ma... 11 main text pages, 10 figures
Aerial World Model for Long-horizon Visual Generation and Navigation in 3D Space	2026-01-03	Show Unmanned aerial vehicles (UAVs) have emerged as powerful embodied agents. One of the core abilities is autonomous navigation in large-scale three-dimensional environments. Existing navigation policies, however, are typically optimized for low-level objectives such as obstacle avoidance and trajectory smoothness, lacking the ability to incorporate high-level semantics into planning. To bridge this gap, we propose ANWM, an aerial navigation world model that predicts future visual observations conditioned on past frames and actions, thereby enabling agents to rank candidate trajectories by their semantic plausibility and navigational utility. ANWM is trained on 4-DoF UAV trajectories and introduces a physics-inspired module: Future Frame Projection (FFP), which projects past frames into future viewpoints to provide coarse geometric priors. This module mitigates representational uncertainty in long-distance visual generation and captures the mapping between 3D trajectories and egocentric observations. Empirical results demonstrate that ANWM significantly outperforms existing world models in long-distance visual forecasting and improves UAV navigation success rates in large-scale environments.
AlignUSER: Human-Aligned LLM Agents via World Models for Recommender System Evaluation	2026-01-02	Show Evaluating recommender systems remains challenging due to the gap between offline metrics and real user behavior, as well as the scarcity of interaction data. Recent work explores large language model (LLM) agents as synthetic users, yet they typically rely on few-shot prompting, which yields a shallow understanding of the environment and limits their ability to faithfully reproduce user actions. We introduce AlignUSER, a framework that learns world-model-driven agents from human interactions. Given rollout sequences of actions and states, we formalize world modeling as a next state prediction task that helps the agent internalize the environment. To align actions with human personas, we generate counterfactual trajectories around demonstrations and prompt the LLM to compare its decisions with human choices, identify suboptimal actions, and extract lessons. The learned policy is then used to drive agent interactions with the recommender system. We evaluate AlignUSER across multiple datasets and demonstrate closer alignment with genuine humans than prior work, both at the micro and macro levels.
NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos	2026-01-01	Show In this paper, we propose NeoVerse, a versatile 4D world model that is capable of 4D reconstruction, novel-trajectory video generation, and rich downstream applications. We first identify a common limitation of scalability in current 4D world modeling methods, caused either by expensive and specialized multi-view 4D data or by cumbersome training pre-processing. In contrast, our NeoVerse is built upon a core philosophy that makes the full pipeline scalable to diverse in-the-wild monocular videos. Specifically, NeoVerse features pose-free feed-forward 4D reconstruction, online monocular degradation pattern simulation, and other well-aligned techniques. These designs empower NeoVerse with versatility and generalization to various domains. Meanwhile, NeoVerse achieves state-of-the-art performance in standard reconstruction and generation benchmarks. Our project page is available at https://neoverse-4d.github.io	Proje... Project Page: https://neoverse-4d.github.io
Device-Native Autonomous Agents for Privacy-Preserving Negotiations	2026-01-01	Show Automated negotiations in insurance and business-to-business (B2B) commerce encounter substantial challenges. Current systems force a trade-off between convenience and privacy by routing sensitive financial data through centralized servers, increasing security risks, and diminishing user trust. This study introduces a device-native autonomous Artificial Intelligence (AI) agent system for privacy-preserving negotiations. The proposed system operates exclusively on user hardware, enabling real-time bargaining while maintaining sensitive constraints locally. It integrates zero-knowledge proofs to ensure privacy and employs distilled world models to support advanced on-device reasoning. The architecture incorporates six technical components within an agentic AI workflow. Agents autonomously plan negotiation strategies, conduct secure multi-party bargaining, and generate cryptographic audit trails without exposing user data to external servers. The system is evaluated in insurance and B2B procurement scenarios across diverse device configurations. Results show an average success rate of 87%, a 2.4x latency improvement over cloud baselines, and strong privacy preservation through zero-knowledge proofs. User studies show 27% higher trust scores when decision trails are available. These findings establish a foundation for trustworthy autonomous agents in privacy-sensitive financial domains.	9 pag... 9 pages, 6 figuers, 9 tables, Submitted in conference 2nd International Conference on Artificial Intelligence Systems (AIS 2026)
YOLO-IOD: Towards Real Time Incremental Object Detection	2026-01-01	Show Current methods for incremental object detection (IOD) primarily rely on Faster R-CNN or DETR series detectors; however, these approaches do not accommodate the real-time YOLO detection frameworks. In this paper, we first identify three primary types of knowledge conflicts that contribute to catastrophic forgetting in YOLO-based incremental detectors: foreground-background confusion, parameter interference, and misaligned knowledge distillation. Subsequently, we introduce YOLO-IOD, a real-time Incremental Object Detection (IOD) framework that is constructed upon the pretrained YOLO-World model, facilitating incremental learning via a stage-wise parameter-efficient fine-tuning process. Specifically, YOLO-IOD encompasses three principal components: 1) Conflict-Aware Pseudo-Label Refinement (CPR), which mitigates the foreground-background confusion by leveraging the confidence levels of pseudo labels and identifying potential objects relevant to future tasks. 2) Importancebased Kernel Selection (IKS), which identifies and updates the pivotal convolution kernels pertinent to the current task during the current learning stage. 3) Cross-Stage Asymmetric Knowledge Distillation (CAKD), which addresses the misaligned knowledge distillation conflict by transmitting the features of the student target detector through the detection heads of both the previous and current teacher detectors, thereby facilitating asymmetric distillation between existing and newly introduced categories. We further introduce LoCo COCO, a more realistic benchmark that eliminates data leakage across stages. Experiments on both conventional and LoCo COCO benchmarks show that YOLO-IOD achieves superior performance with minimal forgetting.	AAAI ... AAAI 2026 accepted. Code & models are available at: https://github.com/qiangzai-lv/YOLO-IOD
TeleWorld: Towards Dynamic Multimodal Synthesis with a 4D World Model	2025-12-31	Show World models aim to endow AI systems with the ability to represent, generate, and interact with dynamic environments in a coherent and temporally consistent manner. While recent video generation models have demonstrated impressive visual quality, they remain limited in real-time interaction, long-horizon consistency, and persistent memory of dynamic scenes, hindering their evolution into practical world models. In this report, we present TeleWorld, a real-time multimodal 4D world modeling framework that unifies video generation, dynamic scene reconstruction, and long-term world memory within a closed-loop system. TeleWorld introduces a novel generation-reconstruction-guidance paradigm, where generated video streams are continuously reconstructed into a dynamic 4D spatio-temporal representation, which in turn guides subsequent generation to maintain spatial, temporal, and physical consistency. To support long-horizon generation with low latency, we employ an autoregressive diffusion-based video model enhanced with Macro-from-Micro Planning (MMPL)--a hierarchical planning method that reduces error accumulation from frame-level to segment-level-alongside efficient Distribution Matching Distillation (DMD), enabling real-time synthesis under practical computational budgets. Our approach achieves seamless integration of dynamic object modeling and static scene representation within a unified 4D framework, advancing world models toward practical, interactive, and computationally accessible systems. Extensive experiments demonstrate that TeleWorld achieves strong performance in both static and dynamic world understanding, long-term consistency, and real-time generation efficiency, positioning it as a practical step toward interactive, memory-enabled world models for multimodal generation and embodied intelligence.
DriveLaW:Unifying Planning and Video Generation in a Latent Driving World	2025-12-31	Show World models have become crucial for autonomous driving, as they learn how scenarios evolve over time to address the long-tail challenges of the real world. However, current approaches relegate world models to limited roles: they operate within ostensibly unified architectures that still keep world prediction and motion planning as decoupled processes. To bridge this gap, we propose DriveLaW, a novel paradigm that unifies video generation and motion planning. By directly injecting the latent representation from its video generator into the planner, DriveLaW ensures inherent consistency between high-fidelity future generation and reliable trajectory planning. Specifically, DriveLaW consists of two core components: DriveLaW-Video, our powerful world model that generates high-fidelity forecasting with expressive latent representations, and DriveLaW-Act, a diffusion planner that generates consistent and reliable trajectories from the latent of DriveLaW-Video, with both components optimized by a three-stage progressive training strategy. The power of our unified paradigm is demonstrated by new state-of-the-art results across both tasks. DriveLaW not only advances video prediction significantly, surpassing best-performing work by 33.3% in FID and 1.8% in FVD, but also achieves a new record on the NAVSIM planning benchmark.	17 pages, 7 figures
What Drives Success in Physical Planning with Joint-Embedding Predictive World Models?	2025-12-30	Show A long-standing challenge in AI is to develop agents capable of solving a wide range of physical tasks and generalizing to new, unseen tasks and environments. A popular recent approach involves training a world model from state-action trajectories and subsequently use it with a planning algorithm to solve new tasks. Planning is commonly performed in the input space, but a recent family of methods has introduced planning algorithms that optimize in the learned representation space of the world model, with the promise that abstracting irrelevant details yields more efficient planning. In this work, we characterize models from this family as JEPA-WMs and investigate the technical choices that make algorithms from this class work. We propose a comprehensive study of several key components with the objective of finding the optimal approach within the family. We conducted experiments using both simulated environments and real-world robotic data, and studied how the model architecture, the training objective, and the planning algorithm affect planning success. We combine our findings to propose a model that outperforms two established baselines, DINO-WM and V-JEPA-2-AC, in both navigation and manipulation tasks. Code, data and checkpoints are available at https://github.com/facebookresearch/jepa-wms.
OSVI-WM: One-Shot Visual Imitation for Unseen Tasks using World-Model-Guided Trajectory Generation	2025-12-30	Show Visual imitation learning enables robotic agents to acquire skills by observing expert demonstration videos. In the one-shot setting, the agent generates a policy after observing a single expert demonstration without additional fine-tuning. Existing approaches typically train and evaluate on the same set of tasks, varying only object configurations, and struggle to generalize to unseen tasks with different semantic or structural requirements. While some recent methods attempt to address this, they exhibit low success rates on hard test tasks that, despite being visually similar to some training tasks, differ in context and require distinct responses. Additionally, most existing methods lack an explicit model of environment dynamics, limiting their ability to reason about future states. To address these limitations, we propose a novel framework for one-shot visual imitation learning via world-model-guided trajectory generation. Given an expert demonstration video and the agent's initial observation, our method leverages a learned world model to predict a sequence of latent states and actions. This latent trajectory is then decoded into physical waypoints that guide the agent's execution. Our method is evaluated on two simulated benchmarks and three real-world robotic platforms, where it consistently outperforms prior approaches, with over 30% improvement in some cases. The code is available at https://github.com/raktimgg/osvi-wm.
World model inspired sarcasm reasoning with large language model agents	2025-12-30	Show Sarcasm understanding is a challenging problem in natural language processing, as it requires capturing the discrepancy between the surface meaning of an utterance and the speaker's intentions as well as the surrounding social context. Although recent advances in deep learning and Large Language Models (LLMs) have substantially improved performance, most existing approaches still rely on black-box predictions of a single model, making it difficult to structurally explain the cognitive factors underlying sarcasm. Moreover, while sarcasm often emerges as a mismatch between semantic evaluation and normative expectations or intentions, frameworks that explicitly decompose and model these components remain limited. In this work, we reformulate sarcasm understanding as a world model inspired reasoning process and propose World Model inspired SArcasm Reasoning (WM-SAR), which decomposes literal meaning, context, normative expectation, and intention into specialized LLM-based agents. The discrepancy between literal evaluation and normative expectation is explicitly quantified as a deterministic inconsistency score, and together with an intention score, these signals are integrated by a lightweight Logistic Regression model to infer the final sarcasm probability. This design leverages the reasoning capability of LLMs while maintaining an interpretable numerical decision structure. Experiments on representative sarcasm detection benchmarks show that WM-SAR consistently outperforms existing deep learning and LLM-based methods. Ablation studies and case analyses further demonstrate that integrating semantic inconsistency and intention reasoning is essential for effective sarcasm detection, achieving both strong performance and high interpretability.
Large Emotional World Model	2025-12-30	Show World Models serve as tools for understanding the current state of the world and predicting its future dynamics, with broad application potential across numerous fields. As a key component of world knowledge, emotion significantly influences human decision-making. While existing Large Language Models (LLMs) have shown preliminary capability in capturing world knowledge, they primarily focus on modeling physical-world regularities and lack systematic exploration of emotional factors. In this paper, we first demonstrate the importance of emotion in understanding the world by showing that removing emotionally relevant information degrades reasoning performance. Inspired by theory of mind, we further propose a Large Emotional World Model (LEWM). Specifically, we construct the Emotion-Why-How (EWH) dataset, which integrates emotion into causal relationships and enables reasoning about why actions occur and how emotions drive future world states. Based on this dataset, LEWM explicitly models emotional states alongside visual observations and actions, allowing the world model to predict both future states and emotional transitions. Experimental results show that LEWM more accurately predicts emotion-driven social behaviors while maintaining comparable performance to general world models on basic tasks.
PhyAVBench: A Challenging Audio Physics-Sensitivity Benchmark for Physically Grounded Text-to-Audio-Video Generation	2025-12-30	Show Text-to-audio-video (T2AV) generation underpins a wide range of applications demanding realistic audio-visual content, including virtual reality, world modeling, gaming, and filmmaking. However, existing T2AV models remain incapable of generating physically plausible sounds, primarily due to their limited understanding of physical principles. To situate current research progress, we present PhyAVBench, a challenging audio physics-sensitivity benchmark designed to systematically evaluate the audio physics grounding capabilities of existing T2AV models. PhyAVBench comprises 1,000 groups of paired text prompts with controlled physical variables that implicitly induce sound variations, enabling a fine-grained assessment of models' sensitivity to changes in underlying acoustic conditions. We term this evaluation paradigm the Audio-Physics Sensitivity Test (APST). Unlike prior benchmarks that primarily focus on audio-video synchronization, PhyAVBench explicitly evaluates models' understanding of the physical mechanisms underlying sound generation, covering 6 major audio physics dimensions, 4 daily scenarios (music, sound effects, speech, and their mix), and 50 fine-grained test points, ranging from fundamental aspects such as sound diffraction to more complex phenomena, e.g., Helmholtz resonance. Each test point consists of multiple groups of paired prompts, where each prompt is grounded by at least 20 newly recorded or collected real-world videos, thereby minimizing the risk of data leakage during model pre-training. Both prompts and videos are iteratively refined through rigorous human-involved error correction and quality control to ensure high quality. We argue that only models with a genuine grasp of audio-related physical principles can generate physically consistent audio-visual content. We hope PhyAVBench will stimulate future progress in this critical yet largely unexplored domain.	6 maj... 6 major physical dimensions, 50 fine-grained test points, 1,000 groups of variable-controlled test samples
Learning to Feel the Future: DreamTacVLA for Contact-Rich Manipulation	2025-12-29	Show Vision-Language-Action (VLA) models have shown remarkable generalization by mapping web-scale knowledge to robotic control, yet they remain blind to physical contact. Consequently, they struggle with contact-rich manipulation tasks that require reasoning about force, texture, and slip. While some approaches incorporate low-dimensional tactile signals, they fail to capture the high-resolution dynamics essential for such interactions. To address this limitation, we introduce DreamTacVLA, a framework that grounds VLA models in contact physics by learning to feel the future. Our model adopts a hierarchical perception scheme in which high-resolution tactile images serve as micro-vision inputs coupled with wrist-camera local vision and third-person macro vision. To reconcile these multi-scale sensory streams, we first train a unified policy with a Hierarchical Spatial Alignment (HSA) loss that aligns tactile tokens with their spatial counterparts in the wrist and third-person views. To further deepen the model's understanding of fine-grained contact dynamics, we finetune the system with a tactile world model that predicts future tactile signals. To mitigate tactile data scarcity and the wear-prone nature of tactile sensors, we construct a hybrid large-scale dataset sourced from both high-fidelity digital twin and real-world experiments. By anticipating upcoming tactile states, DreamTacVLA acquires a rich model of contact physics and conditions its actions on both real observations and imagined consequences. Across contact-rich manipulation tasks, it outperforms state-of-the-art VLA baselines, achieving up to 95% success, highlighting the importance of understanding physical contact for robust, touch-aware robotic agents.
Web World Models	2025-12-29	Show Language agents increasingly require persistent worlds in which they can act, remember, and learn. Existing approaches sit at two extremes: conventional web frameworks provide reliable but fixed contexts backed by databases, while fully generative world models aim for unlimited environments at the expense of controllability and practical engineering. In this work, we introduce the Web World Model (WWM), a middle ground where world state and ``physics'' are implemented in ordinary web code to ensure logical consistency, while large language models generate context, narratives, and high-level decisions on top of this structured latent state. We build a suite of WWMs on a realistic web stack, including an infinite travel atlas grounded in real geography, fictional galaxy explorers, web-scale encyclopedic and narrative worlds, and simulation- and game-like environments. Across these systems, we identify practical design principles for WWMs: separating code-defined rules from model-driven imagination, representing latent state as typed web interfaces, and utilizing deterministic generation to achieve unlimited but structured exploration. Our results suggest that web stacks themselves can serve as a scalable substrate for world models, enabling controllable yet open-ended environments. Project Page: https://github.com/Princeton-AI2-Lab/Web-World-Models.	Proje... Project Page: https://github.com/Princeton-AI2-Lab/Web-World-Models
Act2Goal: From World Model To General Goal-conditioned Policy	2025-12-29	Show Specifying robotic manipulation tasks in a manner that is both expressive and precise remains a central challenge. While visual goals provide a compact and unambiguous task specification, existing goal-conditioned policies often struggle with long-horizon manipulation due to their reliance on single-step action prediction without explicit modeling of task progress. We propose Act2Goal, a general goal-conditioned manipulation policy that integrates a goal-conditioned visual world model with multi-scale temporal control. Given a current observation and a target visual goal, the world model generates a plausible sequence of intermediate visual states that captures long-horizon structure. To translate this visual plan into robust execution, we introduce Multi-Scale Temporal Hashing (MSTH), which decomposes the imagined trajectory into dense proximal frames for fine-grained closed-loop control and sparse distal frames that anchor global task consistency. The policy couples these representations with motor control through end-to-end cross-attention, enabling coherent long-horizon behavior while remaining reactive to local disturbances. Act2Goal achieves strong zero-shot generalization to novel objects, spatial layouts, and environments. We further enable reward-free online adaptation through hindsight goal relabeling with LoRA-based finetuning, allowing rapid autonomous improvement without external supervision. Real-robot experiments demonstrate that Act2Goal improves success rates from 30% to 90% on challenging out-of-distribution tasks within minutes of autonomous interaction, validating that goal-conditioned world models with multi-scale temporal control provide structured guidance necessary for robust long-horizon manipulation. Project page: https://act2goal.github.io/
LidarDM: Generative LiDAR Simulation in a Generated World	2025-12-29	Show We present LidarDM, a novel LiDAR generative model capable of producing realistic, layout-aware, physically plausible, and temporally coherent LiDAR videos. LidarDM stands out with two unprecedented capabilities in LiDAR generative modeling: (i) LiDAR generation guided by driving scenarios, offering significant potential for autonomous driving simulations, and (ii) 4D LiDAR point cloud generation, enabling the creation of realistic and temporally coherent sequences. At the heart of our model is a novel integrated 4D world generation framework. Specifically, we employ latent diffusion models to generate the 3D scene, combine it with dynamic actors to form the underlying 4D world, and subsequently produce realistic sensory observations within this virtual environment. Our experiments indicate that our approach outperforms competing algorithms in realism, temporal coherency, and layout consistency. We additionally show that LidarDM can be used as a generative world model simulator for training and testing perception models.
GaussianDWM: 3D Gaussian Driving World Model for Unified Scene Understanding and Multi-Modal Generation	2025-12-29	Show Driving World Models (DWMs) have been developing rapidly with the advances of generative models. However, existing DWMs lack 3D scene understanding capabilities and can only generate content conditioned on input data, without the ability to interpret or reason about the driving environment. Moreover, current approaches represent 3D spatial information with point cloud or BEV features do not accurately align textual information with the underlying 3D scene. To address these limitations, we propose a novel unified DWM framework based on 3D Gaussian scene representation, which enables both 3D scene understanding and multi-modal scene generation, while also enabling contextual enrichment for understanding and generation tasks. Our approach directly aligns textual information with the 3D scene by embedding rich linguistic features into each Gaussian primitive, thereby achieving early modality alignment. In addition, we design a novel task-aware language-guided sampling strategy that removes redundant 3D Gaussians and injects accurate and compact 3D tokens into LLM. Furthermore, we design a dual-condition multi-modal generation model, where the information captured by our vision-language model is leveraged as a high-level language condition in combination with a low-level image condition, jointly guiding the multi-modal generation process. We conduct comprehensive studies on the nuScenes, and NuInteract datasets to validate the effectiveness of our framework. Our method achieves state-of-the-art performance. We will release the code publicly on GitHub https://github.com/dtc111111/GaussianDWM.
Value-guided action planning with JEPA world models	2025-12-28	Show Building deep learning models that can reason about their environment requires capturing its underlying dynamics. Joint-Embedded Predictive Architectures (JEPA) provide a promising framework to model such dynamics by learning representations and predictors through a self-supervised prediction objective. However, their ability to support effective action planning remains limited. We propose an approach to enhance planning with JEPA world models by shaping their representation space so that the negative goal-conditioned value function for a reaching cost in a given environment is approximated by a distance (or quasi-distance) between state embeddings. We introduce a practical method to enforce this constraint during training and show that it leads to significantly improved planning performance compared to standard JEPA models on simple control tasks.	Prese... Presented as a poster at the World Modeling Workshop 2026, Mila
What Has a Foundation Model Found? Using Inductive Bias to Probe for World Models	2025-12-27	Show Foundation models are premised on the idea that sequence prediction can uncover deeper domain understanding, much like how Kepler's predictions of planetary motion later led to the discovery of Newtonian mechanics. However, evaluating whether these models truly capture deeper structure remains a challenge. We develop a technique for evaluating foundation models that examines how they adapt to synthetic datasets generated from some postulated world model. Our technique measures whether the foundation model's inductive bias aligns with the world model, and so we refer to it as an inductive bias probe. Across multiple domains, we find that foundation models can excel at their training tasks yet fail to develop inductive biases towards the underlying world model when adapted to new tasks. We particularly find that foundation models trained on orbital trajectories consistently fail to apply Newtonian mechanics when adapted to new physics tasks. Further analysis reveals that these models behave as if they develop task-specific heuristics that fail to generalize.	To ap... To appear in ICML 2025
Driving Beyond Privilege: Distilling Dense-Reward Knowledge into Sparse-Reward Policies	2025-12-27	Show We study how to exploit dense simulator-defined rewards in vision-based autonomous driving without inheriting their misalignment with deployment metrics. In realistic simulators such as CARLA, privileged state (e.g., lane geometry, infractions, time-to-collision) can be converted into dense rewards that stabilize and accelerate model-based reinforcement learning, but policies trained directly on these signals often overfit and fail to generalize when evaluated on sparse objectives such as route completion and collision-free overtaking. We propose reward-privileged world model distillation, a two-stage framework in which a teacher DreamerV3-style agent is first trained with a dense privileged reward, and only its latent dynamics are distilled into a student trained solely on sparse task rewards. Teacher and student share the same observation space (semantic bird's-eye-view images); privileged information enters only through the teacher's reward, and the student does not imitate the teacher's actions or value estimates. Instead, the student's world model is regularized to match the teacher's latent dynamics while its policy is learned from scratch on sparse success/failure signals. In CARLA lane-following and overtaking benchmarks, sparse-reward students outperform both dense-reward teachers and sparse-from-scratch baselines. On unseen lane-following routes, reward-privileged distillation improves success by about 23 percent relative to the dense teacher while maintaining comparable or better safety. On overtaking, students retain near-perfect performance on training routes and achieve up to a 27x improvement in success on unseen routes, with improved lane keeping. These results show that dense rewards can be leveraged to learn richer dynamics models while keeping the deployed policy optimized strictly for sparse, deployment-aligned objectives.
InDRiVE: Reward-Free World-Model Pretraining for Autonomous Driving via Latent Disagreement	2025-12-27	Show Model-based reinforcement learning (MBRL) can reduce interaction cost for autonomous driving by learning a predictive world model, but it typically still depends on task-specific rewards that are difficult to design and often brittle under distribution shift. This paper presents InDRiVE, a DreamerV3-style MBRL agent that performs reward-free pretraining in CARLA using only intrinsic motivation derived from latent ensemble disagreement. Disagreement acts as a proxy for epistemic uncertainty and drives the agent toward under-explored driving situations, while an imagination-based actor-critic learns a planner-free exploration policy directly from the learned world model. After intrinsic pretraining, we evaluate zero-shot transfer by freezing all parameters and deploying the pretrained exploration policy in unseen towns and routes. We then study few-shot adaptation by training a task policy with limited extrinsic feedback for downstream objectives (lane following and collision avoidance). Experiments in CARLA across towns, routes, and traffic densities show that disagreement-based pretraining yields stronger zero-shot robustness and robust few-shot collision avoidance under town shift and matched interaction budgets, supporting the use of intrinsic disagreement as a practical reward-free pretraining signal for reusable driving world models.
World Models Unlock Optimal Foraging Strategies in Reinforcement Learning Agents	2025-12-27	Show Patch foraging involves the deliberate and planned process of determining the optimal time to depart from a resource-rich region and investigate potentially more beneficial alternatives. The Marginal Value Theorem (MVT) is frequently used to characterize this process, offering an optimality model for such foraging behaviors. Although this model has been widely used to make predictions in behavioral ecology, discovering the computational mechanisms that facilitate the emergence of optimal patch-foraging decisions in biological foragers remains under investigation. Here, we show that artificial foragers equipped with learned world models naturally converge to MVT-aligned strategies. Using a model-based reinforcement learning agent that acquires a parsimonious predictive representation of its environment, we demonstrate that anticipatory capabilities, rather than reward maximization alone, drive efficient patch-leaving behavior. Compared with standard model-free RL agents, these model-based agents exhibit decision patterns similar to many of their biological counterparts, suggesting that predictive world models can serve as a foundation for more explainable and biologically grounded decision-making in AI systems. Overall, our findings highlight the value of ecological optimality principles for advancing interpretable and adaptive AI.	14 pages, 6 figures
Agent2World: Learning to Generate Symbolic World Models via Adaptive Multi-Agent Feedback	2025-12-26	Show Symbolic world models (e.g., PDDL domains or executable simulators) are central to model-based planning, but training LLMs to generate such world models is limited by the lack of large-scale verifiable supervision. Current approaches rely primarily on static validation methods that fail to catch behavior-level errors arising from interactive execution. In this paper, we propose Agent2World, a tool-augmented multi-agent framework that achieves strong inference-time world-model generation and also serves as a data engine for supervised fine-tuning, by grounding generation in multi-agent feedback. Agent2World follows a three-stage pipeline: (i) A Deep Researcher agent performs knowledge synthesis by web searching to address specification gaps; (ii) A Model Developer agent implements executable world models; And (iii) a specialized Testing Team conducts adaptive unit testing and simulation-based validation. Agent2World demonstrates superior inference-time performance across three benchmarks spanning both Planning Domain Definition Language (PDDL) and executable code representations, achieving consistent state-of-the-art results. Beyond inference, Testing Team serves as an interactive environment for the Model Developer, providing behavior-aware adaptive feedback that yields multi-turn training trajectories. The model fine-tuned on these trajectories substantially improves world-model generation, yielding an average relative gain of 30.95% over the same model before training. Project page: https://agent2world.github.io.	48 pa... 48 pages, 15 tables, 7 figures, Project page: https://agent2world.github.io
Inducing Causal World Models in LLMs for Zero-Shot Physical Reasoning	2025-12-26	Show Large Language Models (LLMs), despite their advanced linguistic capabilities, fundamentally lack an intuitive understanding of physical dynamics, which limits their effectiveness in real-world scenarios that require causal reasoning. In this paper, we introduce Causal World Model Induction (CWMI), a novel framework designed to embed an explicit model of causal physics within an LLM. Our approach incorporates a dedicated Causal Physics Module (CPM) and a new training objective called Causal Intervention Loss, encouraging the model to learn cause-and-effect relationships from multimodal data. By training the model to predict the outcomes of hypothetical interventions instead of merely capturing statistical correlations, CWMI develops a robust internal representation of physical laws. Experimental results show that CWMI significantly outperforms state-of-the-art LLMs on zero-shot physical reasoning tasks, including the PIQA benchmark and our newly proposed PhysiCa-Bench dataset. These findings demonstrate that inducing a causal world model is a critical step toward more reliable and generalizable AI systems.	12 pages, 4 figures,
AstraNav-World: World Model for Foresight Control and Consistency	2025-12-25	Show Embodied navigation in open, dynamic environments demands accurate foresight of how the world will evolve and how actions will unfold over time. We propose AstraNav-World, an end-to-end world model that jointly reasons about future visual states and action sequences within a unified probabilistic framework. Our framework integrates a diffusion-based video generator with a vision-language policy, enabling synchronized rollouts where predicted scenes and planned actions are updated simultaneously. Training optimizes two complementary objectives: generating action-conditioned multi-step visual predictions and deriving trajectories conditioned on those predicted visuals. This bidirectional constraint makes visual predictions executable and keeps decisions grounded in physically consistent, task-relevant futures, mitigating cumulative errors common in decoupled "envision-then-plan" pipelines. Experiments across diverse embodied navigation benchmarks show improved trajectory accuracy and higher success rates. Ablations confirm the necessity of tight vision-action coupling and unified training, with either branch removal degrading both prediction quality and policy reliability. In real-world testing, AstraNav-World demonstrated exceptional zero-shot capabilities, adapting to previously unseen scenarios without any real-world fine-tuning. These results suggest that AstraNav-World captures transferable spatial understanding and planning-relevant navigation dynamics, rather than merely overfitting to simulation-specific data distribution. Overall, by unifying foresight vision and control within a single generative model, we move closer to reliable, interpretable, and general-purpose embodied agents that operate robustly in open-ended real-world settings.
A Unified Definition of Hallucination, Or: It's the World Model, Stupid	2025-12-25	Show Despite numerous attempts to solve the issue of hallucination since the inception of neural language models, it remains a problem in even frontier large language models today. Why is this the case? We walk through definitions of hallucination used in the literature from a historical perspective up to the current day, and fold them into a single definition of hallucination, wherein different prior definitions focus on different aspects of our definition. At its core, we argue that hallucination is simply inaccurate (internal) world modeling, in a form where it is observable to the user (e.g., stating a fact which contradicts a knowledge base, or producing a summary which contradicts a known source). By varying the reference world model as well as the knowledge conflict policy (e.g., knowledge base vs. in-context), we arrive at the different existing definitions of hallucination present in the literature. We argue that this unified view is useful because it forces evaluations to make clear their assumed "world" or source of truth, clarifies what should and should not be called hallucination (as opposed to planning or reward/incentive-related errors), and provides a common language to compare benchmarks and mitigation techniques. Building on this definition, we outline plans for a family of benchmarks in which hallucinations are defined as mismatches with synthetic but fully specified world models in different environments, and sketch out how these benchmarks can use such settings to stress-test and improve the world modeling components of language models.
Motus: A Unified Latent Action World Model	2025-12-25	Show While a general embodied agent must function as a unified system, current methods are built on isolated models for understanding, world modeling, and control. This fragmentation prevents unifying multimodal generative capabilities and hinders learning from large-scale, heterogeneous data. In this paper, we propose Motus, a unified latent action world model that leverages existing general pretrained models and rich, sharable motion information. Motus introduces a Mixture-of-Transformer (MoT) architecture to integrate three experts (i.e., understanding, video generation, and action) and adopts a UniDiffuser-style scheduler to enable flexible switching between different modeling modes (i.e., world models, vision-language-action models, inverse dynamics models, video generation models, and video-action joint prediction models). Motus further leverages the optical flow to learn latent actions and adopts a recipe with three-phase training pipeline and six-layer data pyramid, thereby extracting pixel-level "delta action" and enabling large-scale action pretraining. Experiments show that Motus achieves superior performance against state-of-the-art methods in both simulation (a +15% improvement over X-VLA and a +45% improvement over Pi0.5) and real-world scenarios(improved by +11~48%), demonstrating unified modeling of all functionalities and priors significantly benefits downstream robotic tasks.
Object-Centric World Models for Causality-Aware Reinforcement Learning	2025-12-25	Show World models have been developed to support sample-efficient deep reinforcement learning agents. However, it remains challenging for world models to accurately replicate environments that are high-dimensional, non-stationary, and composed of multiple objects with rich interactions since most world models learn holistic representations of all environmental components. By contrast, humans perceive the environment by decomposing it into discrete objects, facilitating efficient decision-making. Motivated by this insight, we propose \emph{Slot Transformer Imagination with CAusality-aware reinforcement learning} (STICA), a unified framework in which object-centric Transformers serve as the world model and causality-aware policy and value networks. STICA represents each observation as a set of object-centric tokens, together with tokens for the agent action and the resulting reward, enabling the world model to predict token-level dynamics and interactions. The policy and value networks then estimate token-level cause--effect relations and use them in the attention layers, yielding causality-guided decision-making. Experiments on object-rich benchmarks demonstrate that STICA consistently outperforms state-of-the-art agents in both sample efficiency and final performance.	Accepted by AAAI-26
Schrödinger's Navigator: Imagining an Ensemble of Futures for Zero-Shot Object Navigation	2025-12-24	Show Zero-shot object navigation (ZSON) requires a robot to locate a target object in a previously unseen environment without relying on pre-built maps or task-specific training. However, existing ZSON methods often struggle in realistic and cluttered environments, particularly when the scene contains heavy occlusions, unknown risks, or dynamically moving target objects. To address these challenges, we propose \textbf{Schrödinger's Navigator}, a navigation framework inspired by Schrödinger's thought experiment on uncertainty. The framework treats unobserved space as a set of plausible future worlds and reasons over them before acting. Conditioned on egocentric visual inputs and three candidate trajectories, a trajectory-conditioned 3D world model imagines future observations along each path. This enables the agent to see beyond occlusions and anticipate risks in unseen regions without requiring extra detours or dense global mapping. The imagined 3D observations are fused into the navigation map and used to update a value map. These updates guide the policy toward trajectories that avoid occlusions, reduce exposure to uncertain space, and better track moving targets. Experiments on a Go2 quadruped robot across three challenging scenarios, including severe static occlusions, unknown risks, and dynamically moving targets, show that Schrödinger's Navigator consistently outperforms strong ZSON baselines in self-localization, object localization, and overall Success Rate in occlusion-heavy environments. These results demonstrate the effectiveness of trajectory-conditioned 3D imagination in enabling robust zero-shot object navigation.
Active inference and artificial reasoning	2025-12-24	Show This technical note considers the sampling of outcomes that provide the greatest amount of information about the structure of underlying world models. This generalisation furnishes a principled approach to structure learning under a plausible set of generative models or hypotheses. In active inference, policies - i.e., combinations of actions - are selected based on their expected free energy, which comprises expected information gain and value. Information gain corresponds to the KL divergence between predictive posteriors with, and without, the consequences of action. Posteriors over models can be evaluated quickly and efficiently using Bayesian Model Reduction, based upon accumulated posterior beliefs about model parameters. The ensuing information gain can then be used to select actions that disambiguate among alternative models, in the spirit of optimal experimental design. We illustrate this kind of active selection or reasoning using partially observed discrete models; namely, a 'three-ball' paradigm used previously to describe artificial insight and 'aha moments' via (synthetic) introspection or sleep. We focus on the sample efficiency afforded by seeking outcomes that resolve the greatest uncertainty about the world model, under which outcomes are generated.
AD-R1: Closed-Loop Reinforcement Learning for End-to-End Autonomous Driving with Impartial World Models	2025-12-24	Show End-to-end models for autonomous driving hold the promise of learning complex behaviors directly from sensor data, but face critical challenges in safety and handling long-tail events. Reinforcement Learning (RL) offers a promising path to overcome these limitations, yet its success in autonomous driving has been elusive. We identify a fundamental flaw hindering this progress: a deep seated optimistic bias in the world models used for RL. To address this, we introduce a framework for post-training policy refinement built around an Impartial World Model. Our primary contribution is to teach this model to be honest about danger. We achieve this with a novel data synthesis pipeline, Counterfactual Synthesis, which systematically generates a rich curriculum of plausible collisions and off-road events. This transforms the model from a passive scene completer into a veridical forecaster that remains faithful to the causal link between actions and outcomes. We then integrate this Impartial World Model into our closed-loop RL framework, where it serves as an internal critic. During refinement, the agent queries the critic to ``dream" of the outcomes for candidate actions. We demonstrate through extensive experiments, including on a new Risk Foreseeing Benchmark, that our model significantly outperforms baselines in predicting failures. Consequently, when used as a critic, it enables a substantial reduction in safety violations in challenging simulations, proving that teaching a model to dream of danger is a critical step towards building truly safe and intelligent autonomous agents.
FEM-Bench: A Structured Scientific Reasoning Benchmark for Evaluating Code-Generating LLMs	2025-12-23	Show As LLMs advance their reasoning capabilities about the physical world, the absence of rigorous benchmarks for evaluating their ability to generate scientifically valid physical models has become a critical gap. Computational mechanics, which develops and applies mathematical models and numerical methods to predict the behavior of physical systems under forces, deformation, and constraints, provides an ideal foundation for structured scientific reasoning evaluation. Problems follow clear mathematical structure, enforce strict physical and numerical constraints, and support objective verification. The discipline requires constructing explicit models of physical systems and reasoning about geometry, spatial relationships, and material behavior, connecting directly to emerging AI goals in physical reasoning and world modeling. We introduce FEM-Bench, a computational mechanics benchmark designed to evaluate the ability of LLMs to generate correct finite element method (FEM) and related code. FEM-Bench 2025 contains a suite of introductory but nontrivial tasks aligned with material from a first graduate course on computational mechanics. These tasks capture essential numerical and physical modeling challenges while representing only a small fraction of the complexity present in the discipline. Despite their simplicity, state-of-the-art LLMs do not reliably solve all of them. In a five attempt run, the best performing model at function writing, Gemini 3 Pro, completed 30/33 tasks at least once and 26/33 tasks all five times. The best performing model at unit test writing, GPT-5, had an Average Joint Success Rate of 73.8%. Other popular models showed broad performance variation. FEM-Bench establishes a structured foundation for evaluating AI-generated scientific code, and future iterations will incorporate increasingly sophisticated tasks to track progress as models evolve.	40 pa... 40 pages, 5 figures, 6 tables, 7 listings
Active Intelligence in Video Avatars via Closed-loop World Modeling	2025-12-23	Show Current video avatar generation methods excel at identity preservation and motion alignment but lack genuine agency, they cannot autonomously pursue long-term goals through adaptive environmental interaction. We address this by introducing L-IVA (Long-horizon Interactive Visual Avatar), a task and benchmark for evaluating goal-directed planning in stochastic generative environments, and ORCA (Online Reasoning and Cognitive Architecture), the first framework enabling active intelligence in video avatars. ORCA embodies Internal World Model (IWM) capabilities through two key innovations: (1) a closed-loop OTAR cycle (Observe-Think-Act-Reflect) that maintains robust state tracking under generative uncertainty by continuously verifying predicted outcomes against actual generations, and (2) a hierarchical dual-system architecture where System 2 performs strategic reasoning with state prediction while System 1 translates abstract plans into precise, model-specific action captions. By formulating avatar control as a POMDP and implementing continuous belief updating with outcome verification, ORCA enables autonomous multi-step task completion in open-domain scenarios. Extensive experiments demonstrate that ORCA significantly outperforms open-loop and non-reflective baselines in task success rate and behavioral coherence, validating our IWM-inspired design for advancing video avatar intelligence from passive animation to active, goal-oriented behavior.	Proje... Project Page: https://xuanhuahe.github.io/ORCA/
Memorize-and-Generate: Towards Long-Term Consistency in Real-Time Video Generation	2025-12-23	Show Frame-level autoregressive (frame-AR) models have achieved significant progress, enabling real-time video generation comparable to bidirectional diffusion models and serving as a foundation for interactive world models and game engines. However, current approaches in long video generation typically rely on window attention, which naively discards historical context outside the window, leading to catastrophic forgetting and scene inconsistency; conversely, retaining full history incurs prohibitive memory costs. To address this trade-off, we propose Memorize-and-Generate (MAG), a framework that decouples memory compression and frame generation into distinct tasks. Specifically, we train a memory model to compress historical information into a compact KV cache, and a separate generator model to synthesize subsequent frames utilizing this compressed representation. Furthermore, we introduce MAG-Bench to strictly evaluate historical memory retention. Extensive experiments demonstrate that MAG achieves superior historical scene consistency while maintaining competitive performance on standard video generation benchmarks.	Code ... Code will be released at https://github.com/Xilluill/MAG