Skip to content
Sagan

Paper

Physics-R1: An Audited Olympiad Corpus and Recipe for Visual Physics Reasoning

Unreadunread

AI summary

arXiv:2605. 14040v1 Announce Type: new Abstract: We audit the multimodal-physics evaluation pipeline end-to-end and document three undetected construction practices that distort how the field measures vision-language reasoning: train-eval contamination, translation drift, and MCQ saturation.