The authors constructed PhysVista, a benchmark that restores the closed cognitive loop of physical understanding.
New benchmark PhysVista tests vision-language models on full physical cognition loop
The benchmark jointly evaluates perception, reasoning, and plausibility assessment using real and AI-generated videos.
Chinese Tech
Xinge Peng · Yiting Lu · Tianwu Zhi · Wen Wen · Jianzhao Liu · Xin Li · +1 more
University of Science and Technology of China · ByteDance · City University of Hong Kong
Research Digest··2 min read
The authors introduce PhysVista, a benchmark for evaluating physical intelligence in Vision-Language Models (VLMs).
Why this paper
From ByteDance and 2 others
In one line
PhysVista finds that modern vision-language models recognize video content better than they reason about dynamics or judge physical plausibility.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§