Stars
Claude (Anthropic) 账号防风控合规使用 Skill：网络环境、使用规范、内容边界、异常自查与官方申诉指南。
PDF2zh for Zotero | Zotero PDF中文翻译插件
VisRO: procedural visual reasoning tasks, dataset generation, reader and scorer
AutoMMemo: Learning to Learn from Multimodal Experience (NeurIPS 2026)
Trustworthy Recursive Self-Improvement in the Era of Large Models: survey, taxonomy, risks, evaluation, and open challenges.
Complete logs of the experiments in the blog post 'LLMs can understand and write neural network weights'
A Unified Framework for Efficient RSI Research
Investor’s Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?
A curated list of awesome Harbor ecosystem projects
mstar: Optimizing memory architecture for every LLM task as executable Python code.
The official implementation of "EnvScaler: Scaling Tool-Interactive Environments for LLM Agent via Programmatic Synthesis".
A production-grade, high-efficiency Agentic RL framework for frontier LLMs with validated recipes.
🌟100+ 原创 LLM / RL 原理图📚，《大模型算法》作者巨献！💥（100+ LLM/RL Algorithm Maps ）
📚 《从零开始构建智能体》——从零开始的智能体原理与实践教程
A Benchmark for Evaluating Agent Self-Evolution on Real Business Tasks
[ArXiv 2026] Methods for world model interpretability
Official implementation of "Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding"
Reverse engineer anything with agents, from app behavior down to native binaries.
SciForge 科研 Agent 桌面客户端源码（Apache-2.0）：人类控制面与证据层。云端服务与安装器不在此仓。
Terminal-Bench-Science: Evaluating AI agents on research workflows across scientific domains
Turn any AI agent into an AI Scientist. The #1 Agent Skills library for science, used by 250,000+ scientists worldwide. 177 ready-to-use validated skills plus 100+ scientific databases covering bio…
Agent skill that removes signs of AI-generated writing from text
RSIGym: A Flexible Environment for Recursive Self-Improvement
Track 42,331+ AI benchmark, eval, dataset, and data-quality records from 40 public sources, with linked evidence and daily updates.
🧂[Preprint] Salt++: Context-Aligned Post-Training for Few-Step Streaming Multimodal Generation