Reasoning models trained with Group Relative Policy Optimization (GRPO) now produce chains of thought thousands of tokens long, and recent studies reveal tha…
机构:京东
来源:arXiv 2610.34949 | AI4Papers 论文推荐平台