TurnWise Brings Turn-Level Credit to Streaming Video RL
TurnWise is a GRPO framework that assigns reinforcement-learning credit to each decision in a streaming video interaction, rather than sharing one advantage across the full trajectory. The authors say its silence-completed prefix scoring and turn-mean policy optimization improve learned response-versus-wait behavior without requiring a learned critic or extra rollouts, and report outperforming prior RL approaches in extensive experiments.
TurnWise offers a concrete way to align streaming video model training with individual response and waiting decisions, while its reported performance gains come from the authors’ own experiments.
Computer Science > Computer Vision and Pattern Recognition
arXiv:2611.06728 (cs)
[Submitted on 4 Nov 2026]
Title:Turns Matter: Advancing Streaming Video RL via Turn-Level Credit
Authors:Shuai Dong, Kanghui Tian, Linghan Chen, Wenqi Shao, Hongrui Wu, Haowen Hou, Yongfu Zhu, Xuehui Wang, Ziyue Wang, Wenpu Liu, Weichu Xie, Yuqi Xu, Congcong Wang, Kaiwen Tuo, Caoyuan Ma, Taiqiang Wu, Ji Zhao, Wenzheng Chang, Dianyi Wang, Siyuan Wang, Zhongyu Wei, Yu Qiao, Jiaqi Wang
View a PDF of the paper titled Turns Matter: Advancing Streaming Video RL via Turn-Level Credit, by Shuai Dong and 22 other authors