Reinforcement learning (RL) is an effective way to strengthen the long-context reasoning of large language models (LLMs), but it needs training data that are…
机构：清华
来源：arXiv 2610.30132 | AI4Papers 论文推荐平台