Dacheng Tao

MobileSAM2 achieves high-performance segmentation on mobile devices by distilling a heavyweight model into a lightweight version using innovative hypergraph techniques.

Trajectory diversity, not just rollout quantity, is the key to unlocking efficient exploration in VLA reinforcement learning.

Adversarial attacks on vision-language agents reveal critical vulnerabilities, with multi-view optimization strategies proving significantly more effective than isolated approaches.

Error drift in long-horizon video generation can be effectively controlled, leading to state-of-the-art synthesis quality and temporal consistency.

Spatial reasoning could be the secret sauce for building generalist embodied agents that can drive, manipulate objects, and fly drones, all within a single model.