arXiv Computer Vision· Juekai Lin, Honglin Lin, Yuqian Yuan, Xiaolong Wu, Jie Cao, Liang Liang, Yunqi Cao, Yun Zhu, Wenqiao Zhang, Lijun Wu·· 3 小时前AI 评分40
MMVistaReason：面向多模态推理的开放数据与后训练方案
MMVistaReason: Toward Open-Data and Post-Training Recipes for Multimodal Reasoning
AI 导读
MMVistaReason 提出一套开放数据后训练方案，构建 MVR-SFT-528K 与 MVR-RL-63K 数据集，并采用多教师同策略蒸馏（MOPD）整合多个 RL 专家能力。
来源：arXiv Computer Vision · arxiv.org