Baidu（百度）

DuMate-DeepResearch: An Auditable Multi-Agent System with Recursive Search and Rubric-Grounded Reasoning

DuMate-DeepResearch：一种具有递归搜索和基于评分准则推理的可审计多智能体系统

Lingyong Yan, Can Xu, Yukun Zhao, Wenxuan Li, Qingyang Chen, Jiulong Wu, Wenli Song, Xiangnan Li, Weixian Shi, Yiqun Chen, Xuchen Ma, Yuchen Li, Jiashu Zhao, Shuaiqiang Wang, Jianmin Wu, Dawei Yin

机构 * Baidu AI Cloud（百度AI云） 纠错

AI总结 提出DuMate-DeepResearch多智能体框架，通过解耦智能体核心与工具生态、引入图动态规划、递归双层执行和基于评分准则的测试时优化，在深度研究基准上取得最优结果。

Comments Technical report by the DuMate Team. 26 pages, 6 figures, 4 tables

VideoSEG-O3: A Multi-turn Reinforcement Learning Framework for Reasoning Video Object Segmentation

VideoSEG-O3：用于推理视频对象分割的多轮强化学习框架

Ming Dai, Sen Yang, Boqiang Duan, Boyuan Tong, Jiedong Zhuang, Wankou Yang, Jingdong Wang

机构 * School of Automation, Southeast University, Nanjing, China ； Baidu Inc, Beijing, China ； College of Information Science \& Electronic Engineering, Zhejiang University, Hangzhou, China 纠错

AI总结 提出VideoSEG-O3，首个多轮强化学习框架，通过多轮时空思维链和SEG感知逻辑校准，实现从粗到细的推理视频对象分割，解决复杂视频中的精确像素定位问题。

Comments ICML2026