Haoyu Chen

Memory consistency in video generation models falters significantly when objects disappear, with state-of-the-art models struggling to recover updated states upon reappearance.

Stop guessing the right action chunk size for your robot: this method uses action entropy to adaptively determine chunk length, leading to smoother and more responsive manipulation.

Micro-gesture recognition gets a boost from a new method that uses fine-grained semantic guidance to capture subtle motion differences.

By explicitly modeling the shift from local independence to global dependencies between facial action units, micro-AU CLIP achieves state-of-the-art performance in micro-expression analysis.

Finally, a speech emotion dataset that captures *real* spontaneous affect, not just acted emotions, opening the door to more realistic models.