Xiaolong Wang

TactX enables zero-shot transfer of tactile manipulation policies across different sensor types, boosting success rates from 27.5% to 45.9% in complex tasks.

Hallucination in world models can be predicted and mitigated through targeted data collection strategies, transforming how we approach model training in low-coverage scenarios.

Large-scale human motion data can not only train robot controllers but also optimize the physical designs of robot hands, achieving superior performance in real-world applications.

Prioritizing problem difficulty alone can undermine LLM performance, as a structured approach to sampling reveals critical trade-offs in learning efficiency and task coverage.

EgoPhys can predict the physics of unseen deformable objects from just a single egocentric video, revolutionizing how we approach digital twin generation in robotics.

Grounding boosts spatial reasoning in VLMs: explicitly linking language to 2D and 3D scene elements lets models decompose complex spatial problems and improve performance even on non-grounded tasks.

Tactile sensing isn't a universal win for robot manipulation; its effectiveness hinges on the sensor type, material, and the task at hand.