Reproduce “PhyRubric: Rubric-Grounded Diagnostic Evaluation of Physical Commonsense in Text-to-Video Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Huiyi Chen, Weitai Kang, Shicheng Fan, Xinshuo Wang, Zhixiang Shen, Yan Yan, Philip S. Yu, Baoxin Li
PhyRubric evaluates physical commonsense in text-to-video generation by checking caption-specific requirements written before videos are inspected. Its item-level verdicts distinguish missing events from physical violations and attach failures to timestamped evidence. On its benchmark, Seedance 2.5 achieves 44.20% joint performance, versus 14.93% for HunyuanVideo-1.5-8B; this measures benchmark videos, not all possible scenarios.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.