Houqiang Li

4 papers University of Science and Technology of China

• Deep Video Discovery: Agentic Search with Tool Use for Long-form Video Understanding
• Enhancing the Outcome Reward-based RL Training of MLLMs with Self-Consistency Sampling
• Mitigating Hallucination in VideoLLMs via Temporal-Aware Activation Engineering
• Optimizing Distributional Geometry Alignment with Optimal Transport for Generative Dataset Distillation