Watch-Think-Interact: Bootstrapping Long-Horizon Multi-Turn Streaming Video Reasoning with Reinforcement Learning
Watch-Think-Interact (WTI), a closed-loop framework for multi-question streaming video reasoning that maintains compact natural-language memory entries tagged with source-video time ranges, achieves state-of-the-art aggregate performance among the compared open-source streaming baselines.
Zi-Heng Huang, Yi-Cheng Bao, Xue-Heng Li et al.
· 0 citations