Xinhao Li

4 papers University of California, San Diego

• GSPN-2: Efficient Parallel Sequence Modeling
• LongVPO: From Anchored Cues to Self-Reasoning for Long-Form Video Preference Optimization
• StreamForest: Efficient Online Video Understanding with Persistent Event Memory
• VideoChat-R1.5: Visual Test-Time Scaling to Reinforce Multimodal Reasoning by Iterative Perception