Turns Matter: Advancing Streaming Video RL via Turn-Level Credit

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, Taki WU, Ji Zhao, Wenzheng Chang, Dianyi Wang, Siyuan Wang, Zhongyu Wei, Yu Qiao, Jiaqi Wang

Turns Matter: Advancing Streaming Video RL via Turn-Level Credit: 30 upvotes on Hugging Face Daily Papers, #7 of 38 papers on 2026-11-05. Day-by-day upvote history.

Upvotes over time

It had 80 upvotes by its first day, 87 upvotes by the next day, 50 upvotes by day 3, 30 upvotes by day 7, and has 30 now. It ranked #45 of 207 papers in its week and #69 of 340 in its month. Its code repository has 67 GitHub stars.

Streaming video understanding requires models to repeatedly decide whether to respond or continue observing. Recently, Group Relative Policy Optimization (GRPO) has been applied to this setting, with existing approaches emphasizing reward design while retaining a single trajectory-level advantage across turns. Such shared credit can reinforce redundant replies after earlier successes and discourage useful follow-ups after earlier failures. We propose TurnWise, a turn-level GRPO framework that aligns learning with individual streaming decisions. TurnWise addresses the mismatch between trajectory-level feedback and turn-level interaction, promoting informative responses and reducing redundancy. Specifically, turn-level credit assignment uses silence-completed prefix scoring to construct group-relative advantages from each decision's remaining return, separating its learning signal from earlier rewards. Turn-mean policy optimization averages token losses within each turn and then across turns, giving response and waiting decisions equal weight regardless of output length. These complementary designs improve streaming policy learning without a learned critic or additional rollouts. Extensive experiments show our method outperforms prior RL approaches for streaming video understanding.

Paper page on Hugging Face · arXiv

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.