中文版 README

arxiv.abs — arXiv Paper Abstracts for AI Research & Sci-Fi Worldbuilding

48,000+ curated paper abstracts across thirteen collections (1996–2026), covering AI/LLM Prompt Engineering, cross-disciplinary AGI/Sci-Fi research, AI Agent skills, AI-Human futures, health science, game development, Human × AI co-development, software engineering, QA/quality engineering, LLM agent harness research, Anthropic/Claude AI research, child development & education, and US stock market investment analysis.

---

Collections Overview

This repository contains thirteen curated paper abstract collections:

Collection	Papers	Categories	Years
arxiv.prompt	4,780	8 sections	2023–2026
arxiv.agi-scifi	13,606	52 topics	2023–2026
arxiv.agent-skills	2,606	14 skills	2000–2026
arxiv.game-dev	5,197	29 categories	1996–2026
arxiv.ai-human-future	1,814	12 topics	unlimited
arxiv.qbio-health	1,125	7 topics	unlimited
arxiv.game-dev-with-ai	73	25 categories	2013–2026
arxiv.software-engineering	4,418	35 categories	2020–2026
arxiv.qa-quality-engineering	3,069	35 categories	2020–2026
arxiv.agent-harness	2,177	42 topics	2020–2026
arxiv.anthropic-claude	2,628	35 topics	2021–2026
arxiv.child-development	5,072	37 topics	2010–2026
arxiv.us-stock-market	1,241	24 topics	2010–2026
Total	~47,806	345+

---

1. arxiv.prompt — AI/LLM Prompt Engineering

4,780 filtered abstracts + 50 core techniques summary.

Topics: Chain-of-Thought, Few-Shot/ICL, Prompt Optimization, Compression, Jailbreak/Injection, Defense, Multi-Agent, Role-Play, Self-Consistency, and more.

2. arxiv.agi-scifi — Cross-Disciplinary AGI & Sci-Fi Research

13,606 abstracts from 52 keyword searches across all arXiv categories — physics, math, CS, biology, economics, and more.

52 search topics spanning:

Category	Topics
AGI & Intelligence	AGI/Superintelligence, AI Consciousness, Emergence, Scaling Laws, Self-Improving AI, Neuro-Symbolic AI
Safety & Governance	AI Alignment/Safety, AI Deception, AI Governance, AI Military/Weapons, Existential Risk
Society & Economy	AI + Economy/Society, Future of Work, Post-Scarcity, Human-AI Relationship
Physics & Universe	AI + Cosmology, Gravitational Physics, Exotic Physics, Quantum Gravity, Quantum Computing, Theory of Everything
Space & Extraterrestrial	AI + Space, Interstellar Travel, Space Habitat, Terraforming, SETI/Fermi Paradox, Technosignature, Extraterrestrial Life, Extremophile, Origin of Life
Energy & Hardware	AI + Nuclear/Fusion, AI + Climate, AI Hardware/Neuromorphic
Science & Discovery	AI for Science, AI + Biology/Longevity, AI + Scientific Discovery, AI + Math Foundations
Architecture & Systems	World Models, AI Agents, Embodied AI, Collective Intelligence, Decentralized AI, Open-Ended Evolution
Mind & Frontier	Singularity/BCI, Mind Uploading, Simulation/Virtual World, AI Creativity, AI Coding, AI Hallucination, Megastructure/Dyson Sphere
Life Support	Life Support Systems, ISRU

3. arxiv.agent-skills — AI Agent Skills & Capabilities

2,606 abstracts covering 14 core AI agent skill categories: Reasoning, Memory, Tool Use, Planning, Multi-Agent Coordination, Code Agents, RAG, Evaluation, Safety, Embodied AI, Architecture, Self-Improvement, Roleplay, and Long-Horizon Tasks.

4. arxiv.game-dev — Game Development Research (Classical)

5,197 abstracts from 29 categories covering the full history of academic game development research (1996–2026): PCG, game AI, playtesting, game design, graphics, audio, and more.

5. arxiv.ai-human-future — AI & Human Future

1,814 abstracts on 12 topics at the frontier of human-AI coevolution: biological evolution × AI, posthuman futures, mind uploading, brain-computer interfaces, social transformation, and more.

6. arxiv.qbio-health — Quantitative Biology & Health Science

1,125 abstracts on 7 evidence-based health topics: aging & longevity, diet & metabolism, exercise, sleep & circadian biology, gut microbiome, mortality risk factors, and supplements.

7. arxiv.game-dev-with-ai — Human × AI Co-Development (New Era)

73 curated abstracts (filtered from 542 raw) from 25 targeted searches on the emerging field where AI becomes a co-developer and creative partner in game development. Covers LLM-assisted design, co-creative systems, generative assets, autonomous game design, and more.

8. arxiv.software-engineering — Modern Software Engineering (2020+)

4,418 abstracts from 42 keyword searches covering the full modern software engineering stack: desktop/frontend dev, UI/UX, architecture & design patterns, code quality, security, DevOps/CI-CD, AI-assisted development, and software philosophy. Time range: 2020–2026.

9. arxiv.qa-quality-engineering — QA & Quality Engineering (2020+)

3,069 abstracts from 35 keyword searches covering the full QA lifecycle: testing fundamentals, automation frameworks, chaos engineering, AI/ML testing, performance testing, security testing, quality metrics, SRE, and production quality. Aligned with big-tech QE (Quality Engineer) trends. Time range: 2020–2026.

10. arxiv.agent-harness — LLM Agent Research (2020+)

2,177 abstracts from 42 keyword searches covering the full agent lifecycle: frameworks & orchestration, planning & reasoning, tool use, memory, multi-agent coordination, benchmarks, safety, code agents, web/GUI agents, RAG, long-horizon tasks, and observability. Time range: 2020–2026.

11. arxiv.anthropic-claude — Anthropic & Claude Research (2021+)

2,628 abstracts from 35 keyword searches covering Claude-specific and Anthropic research: Constitutional AI, RLHF/RLAIF, mechanistic interpretability, sparse autoencoders, jailbreaking, sycophancy, hallucination, scaling laws, emergent capabilities, and long-context techniques. Time range: 2021–2026.

12. arxiv.child-development — Child Development & Education (2010+)

5,072 abstracts from 37 keyword searches covering child development and education organized by 3-year age brackets (0–18 years), with 80/20 Pareto analysis across 6 developmental layers: attachment/relationship, brain/cognition, social-emotional development, academic outcomes, physical health, and risk/protective factors. Time range: 2010–2026.

13. arxiv.us-stock-market — US Stock Market Investment Analysis (2010+)

1,241 abstracts from 24 keyword searches covering US equity market investment organized by 3-year investment cycle stages, with 80/20 analysis of core factors across 6 analytical layers: macro factors, market timing, stock selection, risk management, behavioral finance, and quantitative strategies. Time range: 2010–2026.

---

Collection Format Standards

All sub-collections follow a unified format standard. See STANDARDS.md for the full spec.

Core requirement: Every category or segment in every collection README must include an 80/20 Pareto best practices summary — identifying the 20% of key practices that drive 80% of outcomes in that domain.

---

Repo Structure

arxiv.abs/
├── README.md                              # English
├── README.zh-CN.md                        # Chinese
├── STANDARDS.md                           # Collection format standards
├── arxiv.prompt/                          # Prompt engineering (4,780 papers)
│   ├── README.md / README.zh-CN.md
│   ├── prompt_ai.md / prompt_ai.zh-CN.md  # 50 core techniques
│   └── 2023–2026/abstracts/
├── arxiv.agi-scifi/                       # AGI & sci-fi (13,606 papers)
│   ├── README.md / README.zh-CN.md
│   ├── agi-scifi.md / agi-scifi.zh-CN.md  # Key findings
│   ├── search-index.json
│   └── 2023–2026/abstracts/
├── arxiv.agent-skills/                    # Agent skills (2,606 papers)
│   ├── README.md / README.zh-CN.md
│   ├── agent-skills-findings.zh-CN.md
│   ├── search-index.json
│   └── 2000–2026/abstracts/
├── arxiv.game-dev/                        # Game dev classical (5,197 papers)
│   ├── README.md / README.zh-CN.md
│   ├── readme_game_roadmap.md
│   ├── search-index.json
│   └── 1996–2026/abstracts/
├── arxiv.ai-human-future/                 # AI-human futures (1,814 papers)
│   ├── README.md / README.zh-CN.md
│   ├── ai-future-findings.zh-CN.md
│   ├── search-index.json
│   └── abstracts/
├── arxiv.qbio-health/                     # Health science (1,125 papers)
│   ├── README.md / README.zh-CN.md
│   ├── health-findings.zh-CN.md
│   ├── search-index.json
│   └── abstracts/
└── arxiv.game-dev-with-ai/                # Human × AI co-dev (73 curated papers)
    ├── README.md / README.zh-CN.md
    ├── readme_game_dev_with_ai.md          # Landscape & practitioner roadmap
    ├── search-index.json
    └── 2005–2026/abstracts/

---

Quick Start

Browse the prompt techniques:

Start with arxiv.prompt/prompt_ai.md — 50 techniques with templates, principles, sources, and use cases.

Explore sci-fi research:

Browse the arxiv.agi-scifi/ directory by year — each abstract is a self-contained .txt file with title, authors, arXiv ID, and abstract text.

Programmatic access:

import json

data = json.load(open("arxiv.agi-scifi/2025/arxiv-results-2025.json"))
for paper in data["papers"]:
    print(paper["title"], paper["categories"], paper.get("citations", "N/A"))

---

Methodology

Prompt Collection

1. Searched arXiv API for CS papers with "prompt" in title (2023–2026)
2. Retrieved 5,362 abstracts via arxs CLI tool
3. Filtered out 582 non-LLM papers (vision, speech, graph, diffusion, medical)
4. Categorized 4,780 papers and distilled 50 core techniques

AGI/Sci-Fi Collection

1. Designed 52 cross-disciplinary keyword searches across all arXiv categories
2. Retrieved 14,520 raw results covering physics, math, CS, biology, economics, etc.
3. Deduplicated to 13,606 unique papers
4. Organized by year with formatted abstracts (80 chars/line)

---

50 Prompt Engineering Techniques

Full details in arxiv.prompt/prompt_ai.md:

Section	Techniques	Range
I. Reasoning & CoT	CoT, Plan-and-Solve, Structured CoT, etc.	#1–6
II. Examples & ICL	Few-Shot, Auto-Demo, Self-Consistency, etc.	#7–11
III. Design & Formatting	Pattern Catalog, Role-Play, Emotional Stimulus, etc.	#12–16
IV. Orchestration	Chaining, Meta-Prompting, Multi-Agent, etc.	#17–22
V. Optimization	Auto Optimization, Compression, Soft Prompt, etc.	#23–28
VI. Quality & Robustness	Bias Calibration, Sensitivity, LLM-as-Judge, etc.	#29–33
VII. Defense	Spotlighting, Query Defense, Injection Detection, etc.	#34–41
VIII. Attacks (Research)	Jailbreak, Prompt Injection, Extraction, etc.	#42–50

---

52 AGI/Sci-Fi Search Keywords

Click to expand full search list

#	Search Topic	Papers
01	AGI / Superintelligence	434
02	AI + Economy / Society	500
03	AI + Cosmology / Universe	500
04	AI Consciousness	51
05	Human-AI Relationship	500
06	AI + Nuclear Energy	195
07	AI Safety / Alignment	500
08	AI Governance	423
09	AI + Space Exploration	500
10	AI + Math Foundations	26
11	AI + Quantum Computing	500
12	Singularity / BCI	239
13	Emergence in AI	500
14	AI + Military / Weapons	22
15	World Models	500
16	Self-Improving AI	268
17	AI + Climate / Energy	500
18	AI + Biology / Longevity	460
19	Collective Intelligence	500
20	Simulation / Virtual World	500
21	Neuro-Symbolic AI	500
22	AI + Scientific Discovery	261
23	Decentralized AI	252
24	AI + Creativity / Art	365
25	AI Autonomous Agents	500
26	Embodied AI	500
27	Megastructure / Dyson Sphere	20
28	AI for Science	194
29	SETI / Fermi Paradox	57
30	Information Theory + Intelligence	500
31	Future of Work	500
32	Mind Uploading	18
33	Scaling Laws	500
34	AI Deception / Misalignment	63
35	AI + Gravitational Physics	302
36	Post-Scarcity Economics	7
37	Theory of Everything	39
38	AI Hardware / Neuromorphic	500
39	AI + Exotic Physics	124
40	AI + Fusion Control	44
41	AI Code Generation	355
42	AI Hallucination / Trust	307
43	Open-Ended Evolution	189
44	Quantum Gravity + Information	322
45	Extraterrestrial Life / Astrobiology	199
46	Space Habitat / Colonization	83
47	Life Support Systems / ISRU	9
48	Extremophile / Space Biology	54
49	Interstellar Travel / Propulsion	55
50	Terraforming	11
51	Origin of Life / Panspermia	20
52	Technosignature / Cosmic Archaeology	52

---

Use Cases

• Prompt engineers — find and apply the latest LLM prompting techniques
• AI safety researchers — study alignment, deception, jailbreak, and defense
• Sci-fi writers — discover real scientific ideas for worldbuilding
• Students & researchers — quickly survey cross-disciplinary AI literature

---

Contributing

Issues and PRs welcome!

• Suggest missing topics or techniques
• Correct categorizations
• Add new search keywords
• Improve summaries

---

License

Abstracts are sourced from arXiv under their Terms of Use. Thank you to arXiv for use of its open access interoperability.

---

If this repo helps your research or work, please star it! ⭐