LFHF.md

• Learning to summarize from human feedback

OpenAI采用LFHF技术在NLP领域的初步尝试

• Raining a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

Anthropic团队尝试用LFHF技术解决Harmless Assistant问题

• InstructGPT: Training language models to follow instructions with human feedback

OpenAI采用LFHF技术在NLP领域的进一步尝试，也是ChatGPT的前身

• Constitutional AI: Harmlessness from AI Feedback

针对Human feedback效率低的问题，提出AI feedback方案

• Scaling Laws for Reward Model Overoptimization

分析 Reward Model 细节的文章