Feng Xue

Adversarial training doesn't have to hurt speaker verification: by explicitly modeling language, you can disentangle speaker and language characteristics without sacrificing speaker discriminability.

Unlock efficient robot locomotion on granular terrains by inferring terrain properties from arbitrary gaits, enabling active adaptation and optimized foot design.

End-to-end autonomous driving can ditch expert demonstrations and still achieve state-of-the-art performance, thanks to a risk-aware world model that learns to predict and avoid hazardous outcomes.