benchmark · first seen 24 Oct 2026
physical reasoning benchmark
Evaluation suite testing whether models reason about physics from visual evidence under altered laws.
Mentions per week
PapersCodeDiscussion
Earliest sources
- InfiniteWorld: Separating Visual Evidence from Familiar PhysicsPapers · 24 Oct 2026 · Tong Zhang, Yuxuan Cheng, Yun Peng
Latest sources
- InfiniteWorld: Separating Visual Evidence from Familiar PhysicsPapers · 24 Oct 2026 · Tong Zhang, Yuxuan Cheng, Yun Peng