Research

Paper

AI LLM February 23, 2026

Pixel2Phys: Distilling Governing Laws from Visual Dynamics

Authors

Ruikun Li, Jun Yao, Yingfan Hua, Shixiang Tang, Biqing Qi, Bin Liu, Wanli Ouyang, Yan Lu

Abstract

Discovering physical laws directly from high-dimensional visual data is a long-standing human pursuit but remains a formidable challenge for machines, representing a fundamental goal of scientific intelligence. This task is inherently difficult because physical knowledge is low-dimensional and structured, whereas raw video observations are high-dimensional and redundant, with most pixels carrying little or no physical meaning. Extracting concise, physically relevant variables from such noisy data remains a key obstacle. To address this, we propose Pixel2Phys, a collaborative multi-agent framework adaptable to any Multimodal Large Language Model (MLLM). It emulates human scientific reasoning by employing a structured workflow to extract formalized physical knowledge through iterative hypothesis generation, validation, and refinement. By repeatedly formulating, and refining candidate equations on high-dimensional data, it identifies the most concise representations that best capture the underlying physical evolution. This automated exploration mimics the iterative workflow of human scientists, enabling AI to reveal interpretable governing equations directly from raw observations. Across diverse simulated and real-world physics videos, Pixel2Phys discovers accurate, interpretable governing equations and maintaining stable long-term extrapolation where baselines rapidly diverge.

Metadata

arXiv ID: 2602.19516
Provider: ARXIV
Primary Category: cs.CE
Published: 2026-02-23
Fetched: 2026-02-24 04:38

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.19516v1</id>\n    <title>Pixel2Phys: Distilling Governing Laws from Visual Dynamics</title>\n    <updated>2026-02-23T05:16:47Z</updated>\n    <link href='https://arxiv.org/abs/2602.19516v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.19516v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Discovering physical laws directly from high-dimensional visual data is a long-standing human pursuit but remains a formidable challenge for machines, representing a fundamental goal of scientific intelligence. This task is inherently difficult because physical knowledge is low-dimensional and structured, whereas raw video observations are high-dimensional and redundant, with most pixels carrying little or no physical meaning. Extracting concise, physically relevant variables from such noisy data remains a key obstacle. To address this, we propose Pixel2Phys, a collaborative multi-agent framework adaptable to any Multimodal Large Language Model (MLLM). It emulates human scientific reasoning by employing a structured workflow to extract formalized physical knowledge through iterative hypothesis generation, validation, and refinement. By repeatedly formulating, and refining candidate equations on high-dimensional data, it identifies the most concise representations that best capture the underlying physical evolution. This automated exploration mimics the iterative workflow of human scientists, enabling AI to reveal interpretable governing equations directly from raw observations. Across diverse simulated and real-world physics videos, Pixel2Phys discovers accurate, interpretable governing equations and maintaining stable long-term extrapolation where baselines rapidly diverge.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CE'/>\n    <published>2026-02-23T05:16:47Z</published>\n    <arxiv:comment>CVPR2026 main track</arxiv:comment>\n    <arxiv:primary_category term='cs.CE'/>\n    <author>\n      <name>Ruikun Li</name>\n    </author>\n    <author>\n      <name>Jun Yao</name>\n    </author>\n    <author>\n      <name>Yingfan Hua</name>\n    </author>\n    <author>\n      <name>Shixiang Tang</name>\n    </author>\n    <author>\n      <name>Biqing Qi</name>\n    </author>\n    <author>\n      <name>Bin Liu</name>\n    </author>\n    <author>\n      <name>Wanli Ouyang</name>\n    </author>\n    <author>\n      <name>Yan Lu</name>\n    </author>\n  </entry>"
}