Reinforcement learning with verifiable rewards (RLVR) has significantly improved the reasoning capabilities of large vision-language models (LVLMs). However,…
机构:清华
来源:arXiv 2609.18057 | AI4Papers 论文推荐平台