SeePhys: Does Seeing Help Thinking? -- Benchmarking Vision-Based Physics Reasoning

TL;DR

SeePhys is a comprehensive benchmark testing vision-based physics reasoning in large language models, revealing significant challenges in visual understanding and reasoning integration.

Contribution

The paper introduces SeePhys, a large-scale, diverse benchmark emphasizing vision-essential physics problems to evaluate and challenge current multimodal reasoning models.

Findings

Most advanced models score below 60% accuracy.

Visual reasoning remains a significant challenge for LLMs.

Models tend to rely on textual cues rather than visual understanding.

Abstract

We present SeePhys, a large-scale multimodal benchmark for LLM reasoning grounded in physics questions ranging from middle school to PhD qualifying exams. The benchmark covers 7 fundamental domains spanning the physics discipline, incorporating 21 categories of highly heterogeneous diagrams. In contrast to prior works where visual elements mainly serve auxiliary purposes, our benchmark features a substantial proportion of vision-essential problems (75%) that mandate visual information extraction for correct solutions. Through extensive evaluation, we observe that even the most advanced visual reasoning models (e.g., Gemini-2.5-pro and o4-mini) achieve sub-60% accuracy on our benchmark. These results reveal fundamental challenges in current large language models' visual understanding capabilities, particularly in: (i) establishing rigorous coupling between diagram interpretation and physics reasoning, and (ii) overcoming their persistent reliance on textual cues as cognitive shortcuts.