Jiahui Fu

TopoGPT achieves a remarkable leap in lane topology reasoning, producing geometrically consistent lane graphs that outperform existing methods by substantial margins.

Pruning 90% of visual tokens without sacrificing performance could revolutionize the efficiency of 3D scene understanding in multimodal models.

Robots can now perform intricate assembly tasks and recover from errors in real-time, without any training, by fusing vision-language models with video-based kinematic priors for action planning.