Learning to Watch: Active Video Anomaly Understanding via Interleaved Policy Optimization
学习观察:通过交错策略优化的主动视频异常理解
机构 * School of Computer Science and Technology, Chongqing University of Posts and Telecommunications, Chongqing, China(重庆邮电大学计算机科学与技术学院) ; School of Computer Science(计算机科学学院) ; Chongqing College of Artificial Intelligence, Chongqing, China(重庆人工智能学院)
专题命中 规划决策 :agent(abstract)
AI总结 提出Anom-π闭环框架,将视频异常理解建模为主动序列决策过程,通过交错策略统一内部推理与证据获取,并设计iDPO实现轨迹级策略对齐,仅2B参数即超越大规模模型。
Comments Accepted at ICML 2026; 25 pages, 8 figures, 15 tables