Large Reasoning Models (LRMs) generate long reasoning traces that expose intermediate inference behavior, offering opportunities to study and control reasoni…
机构:上交
来源:arXiv 2610.34943 | AI4Papers 论文推荐平台