Principal Researcher, Xiaomi Corporation
3 papers at NeurIPS 2025
We present Time-R1, a reinforcement learning-based post-training framework that achieves state-of-the-art performance across large vision language models for temporal video grounding.
