Kunchang Li

Make your training flexible: towards deployment-efficient video models

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

TimeSuite: Improving MLLMs for Long Video Understanding via Grounded Tuning