让AI自己给自己“立规矩”，结果会怎样？
[LG] AutoRule: Reasoning Chain-of-thought Extracted Rule-based Rewards Improve Preference Learning
[CMU]
https://arxiv.org/abs/2506.15651
在小宇宙查看该单集文稿
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.