Postdoc, University of California, Berkeley
1 paper at NeurIPS 2025
This paper introduces the first unified platform for evaluating text-to-video and video-to-text models across five key dimensions: safety, hallucination, fairness, privacy, and adversarial robustness.
