Learning to solve hard problems in RL for LLMs by never giving up

87 points by natolambert 13 hours ago on hackernews | 4 comments
https://arxiv.org/abs/2609.13443