Reward Model and RLHF

课程大纲

• 大模型评估 [课件]
• 对齐和RLHF [课件]

推荐阅读材料

• [论文]Training language models to follow instructions with human feedback
• [论文]Direct Preference Optimization: Your Language Model is Secretly a Reward Model
• [论文]BARTScore: Evaluating Generated Text as Text Generation
• [论文]GPTScore: Evaluate as You Desire
• [论文]Can Large Language Models be Trusted for Evaluation? Scalable Meta-Evaluation of LLMs as Evaluators via Agent Debate

致谢

• 感谢 Yixiu Liu 协助一起完成指令学习的课件。