Ning Mao

Efficient dense 3D reconstruction can now serve as a scalable foundation for next-generation autonomous driving, achieving real-time performance without sacrificing fidelity.

Tiny VLMs can punch far above their weight: a 0.5B parameter model, guided by Switch-KD, closes the gap with its 3B teacher by 3.6 points on multimodal benchmarks.

KANs don't just predict crystal properties with SOTA accuracy; they also reveal the underlying chemistry, offering a glimpse into the model's reasoning.

A breast pathology foundation model, BRIGHT, beats generalist models at specialist tasks, suggesting organ-specific pretraining is a scalable path to expert-level performance.