arXiv · 机器人与计算机视觉· Xuan Gong, Hanbo Huang, Hao Zheng, Yiran Zhang, Wenbin Dai, Weishu Zhao, Shiyu Liang·· 2 天前AI 评分82
Reflection Anchors for Propagation-Aware Visual Retention in Long-Chain Multimodal Reasoning
Reflection Anchors for Propagation-Aware Visual Retention in Long-Chain Multimodal Reasoning
AI 导读
Xuan Gong 等人提出 RAPO(Reflection Anchor Policy Optimization),一种基于 GRPO 的策略优化方法,旨在解决长链思维推理中视觉信息衰减问题。
来源:arXiv · 机器人与计算机视觉 · arxiv.org