AIPaperNotes

Record daily reading of papers and related reproduction results in Chinese.

For more notes, please follow the blog: https://nopsled.blog.csdn.net/

Paper Catalog

Embedding

• NV-EMBED [Link]

Flow

• Flow Matching [Link]

LLM

• Agent: LLM-based Single/Multi Agent model/system

  • DeepResearch:
    • WebSailor [Link]
    • WebWatcher [Link]
  • Memory:
    • Dynamic Cheatsheet [Link]
    • EgoMem [Link]
  • Reflection:
    • Metacognitive Reuse [Link]
  • Router:
    • Router-R1 [Link]
  • Visual Reasoning
    • PixelCraft [Link]

• Base Model: Large Language Model

  • Zhipu AI
    • GLM 4.5 [Link]
  • OpenAI
    • gpt-oss [Link]

• Dataset: Data building and processing for Model training

  • DELT [Link]

• Prompt: Prompt Engineering

  • Context Learning
    • ACE [Link]

• Omni: LLM-based full modal model

  • Qwen2.5 - Omni [Link]
  • M3 - Agent [Link]

• Quantization: Model Weight/Optimizer/Activation Compressing

  • COAT [Link]

• Speech: Speech LLM

  • ALM: Audio LLM for auido Input
    • Audio Flamingo 3 [Link]

• Survey

  • DeepRearch [Link]
  • Vibe Coding

• Training: LLM Model Training:

  • Long CoT
    • Deepseek - R1 [Link]
    • Dr.GRPO [Link]
    • DAPO [Link]
    • GCG [Link]
    • LUFFY [Link]
  • Ptrtrain
    • FIM (fill-in-the-middle) [Link]
  • RL
    • BCO [Link]
  • Speculative Decoding or MTP: Speculative Decoding or Multi-token Prediction
    • Better & Faster Large Language Models via Multi-token Prediction [Link]
    • CAFT [Link]
    • EAGLE3 [Link]

• VLM: Model Weight/Optimizer/Activation Compressing

  • Qwen - VL [Link]
  • LLaVA [Link]
  • Qwen2 - VL [Link]
  • Qwen2.5 - VL
  • MiniCPM-V 4.5 [Link]
  • DeepSeek - OCR [Link]

Visual Encoder

• Language-Image Representation Learning:
  • CLIP [Link]
  • SigLIP [Link]
  • LIFT [Link]