Sijia Chen

Autonomous driving planners can now explicitly self-correct unsafe actions by generating motion-token traces conditioned on a learned collision critic, leading to significant safety improvements.

Omnidirectional imagery + language unlocks robust multi-object tracking that overcomes the field-of-view limitations plaguing conventional video datasets.

Now you can track multiple objects in low-visibility conditions using language descriptions, thanks to a new RGB-Thermal dataset and MLLM-based framework.