Xiangyu Zeng

TimeSuite: Improving MLLMs for Long Video Understanding via Grounded Tuning

Task preference optimization: improving multimodal large language models with vision task alignment

Online video understanding: a comprehensive benchmark and memory-augmented method