PhysUniBench: A Multi-Modal Physics Reasoning Benchmark at Undergraduate Level

TL;DR

PhysUniBench addresses the need for a rigorous undergraduate-level, multimodal physics benchmark by aggregating $3{,}304$ diagram-supported questions across eight subfields, with open-ended and multiple-choice formats and bilingual EN/ZH availability. The authors implement a three-phase data-curation pipeline to ensure difficulty stratification (levels $1$–$5$) and high-quality items, including AI-powered answer generation, filtering, and MC conversion using the model's own failure modes. Extensive zero-shot evaluations of state-of-the-art multimodal large language models reveal substantial challenges in integrating physics knowledge, symbolic reasoning, and diagram interpretation, with GPT-5.2 delivering the best open-ended score ($59.7\%$) and GPT-5 the best multiple-choice score ($63.6\%$), while performance declines sharply at higher difficulty levels. The benchmark thus serves as a diagnostic, curriculum-aligned testbed to guide future advances in multimodal physics reasoning and diagrammatic understanding, with implications for AI-assisted science education and assessment.

Abstract

Physics problem-solving is a challenging domain for AI models, requiring integration of conceptual understanding, mathematical reasoning, and interpretation of physical diagrams. Existing evaluations fail to capture the full breadth and complexity of undergraduate physics, whereas this level provides a rigorous yet standardized testbed for pedagogical assessment of multi-step physical reasoning. To this end, we present PhysUniBench, a large-scale multimodal benchmark designed to evaluate and improve the reasoning capabilities of multimodal large language models (MLLMs) specifically on undergraduate-level physics problems. PhysUniBench consists of 3,304 physics questions spanning 8 major sub-disciplines of physics, each accompanied by one visual diagram. The benchmark includes both open-ended and multiple-choice questions, systematically curated and difficulty-rated through an iterative process. The benchmark's construction involved a rigorous multi-stage process, including multiple roll-outs, expert-level evaluation, automated filtering of easily solved problems, and a nuanced difficulty grading system with five levels. Through extensive experiments, we observe that current models encounter substantial challenges in physics reasoning, where GPT-5 achieves only 51.6% accuracy in the PhysUniBench. These results highlight that current MLLMs struggle with advanced physics reasoning, especially on multi-step problems and those requiring precise diagram interpretation. By providing a broad and rigorous assessment tool, PhysUniBench aims to drive progress in AI for Science, encouraging the development of models with stronger physical reasoning, problem-solving skills, and multimodal understanding.

Figures (14)

• Figure 1: PhysUniBench is the first large-scale multimodal benchmark designed for evaluating undergraduate-level physics understanding, reasoning, and problem-solving. It includes 3,304 rigorously curated questions from authentic university curricula.
• Figure 2: PhysUniBench is constructed through a rigorous three-stage data curation process designed to ensure high-quality multimodal physics problems. This pipeline systematically curates a wide range of questions across 8 core physics disciplines.
• Figure 3: Performance of Gemini-2.5-Pro with caption input.
• Figure 4: Four error types on OE of mechanics sub-discipline.
• Figure 5: Diverse Distribution of PhysUniBench.