// the find
WangRongsheng/awesome-LLM-resources
🧑🚀 全世界最好的LLM资料总结（多模态生成、Agent、辅助编程、AI审稿、数据处理、模型训练、模型推理、o1 模型、MCP、小语言模型、视觉语言模型） | Summary of the world's best LLM resources.
A large Chinese-maintained awesome list aggregating LLM tools, papers, courses, and resources across the full ML stack — data processing, fine-tuning, inference, RAG, agents, evaluation, and more. Aimed at Chinese-speaking ML practitioners building or studying LLMs, though many linked projects are English. Updated daily.
Covers the full LLM development lifecycle in one place, from data labeling (MinerU, data-juicer) through fine-tuning (LLaMA-Factory, veRL) to serving (vLLM, SGLang) — useful as a starting index when you don't know what exists. Active maintenance: last push was yesterday and new categories (Agentic RL, World Models, Unified Models) track the field's current obsessions. Includes actual PDF books and technical reports (DeepSeek R1/V3, Kimi k1.5) checked directly into the repo — convenient, even if the copyright situation is questionable. The 🔥 and 🌟 markers give a rough signal of which entries the maintainer considers most important.
No curation — this is a link dump, not a guide. There are duplicate entries (OpenRLHF, LightRAG, KAG each appear twice in the same section) and no indication of when something was added or why it's better than the alternatives next to it. The README is primarily Chinese; section names are bilingual but all commentary, podcast summaries, and notes are Chinese-only, which makes it effectively inaccessible to non-Chinese readers despite linking to mostly English projects. Committing PDFs of commercial technical reports and published books into a public repo is a copyright problem waiting to happen. At this size, the list has the classic awesome-list problem: it tells you everything exists and helps you decide nothing.