DecryptPrompt
A continuously updated Chinese-language collection of summarized prompt-engineering and LLM papers, open datasets and AIGC applications, with an accompanying blog series.
DecryptPrompt is a continuously updated resource collection covering large language models and prompt engineering: open-source models and evaluation leaderboards, open-source inference, fine-tuning, agent and RAG frameworks, open SFT, RLHF and pretraining datasets, and a survey of AIGC applications across domains.
Alongside the collection, the author writes a numbered blog series, in Chinese, that works through the material topic by topic, covering areas such as prompt tuning methods, instruction tuning, chain-of-thought reasoning, retrieval-augmented generation, LLM agents, reinforcement learning from human and AI feedback, and reasoning-model training techniques such as those behind DeepSeek R1 and Kimi 1.5.
Its GitHub topics include prompt engineering, in-context learning, few-shot and zero-shot learning, chain-of-thought and instruction tuning.