Chen Cai

A single differentiable model now predicts inter-arm collision distances for real-time, gradient-based multi-robot motion planning, overcoming the limitations of traditional geometric checkers and conservative proxies.

Timestamp-supervised video segmentation consistently fails at transition boundaries, but aggregating video-wide keyframe "votes" resolves local ambiguity without requiring dense manual annotations.

Visual agents consistently falter at detecting procedural errors, but a new framework shows how to enhance their comprehension capabilities dramatically.

Real-robot trials are costly and slow, but DataLadder enables scalable evaluations and data generation through a seamless interplay between robots, simulations, and human feedback.

Adaptive expert collaboration and holistic token learning can significantly boost performance in fine-grained multimodal visual analytics tasks like driver action recognition.