Multimodal reinforcement learning with verifiable rewards (RLVR) provides reliable task-level supervision, yet offers limited guidance for localizing errors …
机构:百度
来源:arXiv 2610.26856 | AI4Papers 论文推荐平台