Desen Meng

LongVPO: from anchored cues to self-reasoning for long-form video preference optimization

Oct 11, 2025

p-MoD: building mixture-of-depths MLLMs via progressive ratio decay

Aug 12, 2025