PhysUniBench
Evaluating multimodal models on undergraduate physics through diagram-grounded questions and scientific reasoning.
PhysUniBench evaluates whether multimodal models can combine visual diagrams, physical concepts, and mathematical reasoning to solve undergraduate-level physics problems. It covers eight physics subfields, from classical mechanics and electromagnetism to quantum mechanics and relativity.
The benchmark includes both open-ended and multiple-choice questions. Its visual evidence and difficulty annotations support analysis of scientific reasoning across topics and levels of complexity.