Jul 2026
TimeThink: Reasoning with Time for Video LLMs
TimeThink is proposed, a reinforcement learning framework that explicitly guides temporal evidence discovery in Video-LLMs and introduces a step-wise temporal process reward that provides localized credit assignment for these clues and a joint process--outcome optimization objective that balances reasoning fidelity with task correctness.
Handong Li, Longteng Guo, Zikang Liu et al.
· arXiv.org · 0 citations