PhysElite: How Far Are LLMs from Solving Olympiad-Level Physics Problems?
Abstract
Understanding how (multimodal) large language models perform on physics problems requires benchmarks that reflect the difficulty and breadth of expert-level physical reasoning. Existing physics benchmarks remain limited in the following two important ways: (1) short of high-difficulty datasets, and (2) lack of comprehensive coverage of visual forms, knowledge points, and step-by-step solution processes. As a result, model performance on current datasets may not be fully representative of their ability to solve complex physics problems. To address these issues, we present PhysElite, a large-scale bilingual multimodal benchmark for Olympiad-level physics reasoning. PhysElite contains 11,586 Olympiad-tier problems. For each problem, we provide corresponding visual diagrams, step-by-step bilingual Chinese-English solution derivations, and the final answer. We benchmark 18 open-source and closed-source MLLMs, and find that even the strongest model reaches only 33.7% answer accuracy. We additionally conduct step-level process evaluation to diagnose where models fail in the reasoning chain. Our datasets are released at https://huggingface.co/datasets/physelite/PhysElite.
Related
- A Framework Integrating the Dynamic Stiffness Matrix with Physics-Informed Neural Networks for Solving Eigenvalue Problems and Analysing Dynamic Response
- Solving Inverse Dirac-weighted Sturm-Liouville Problems via Cauchy problems
- Solving Multiobjective Linear Set Optimization Problems
- Error Analysis for Solving Elliptic Interface Problems with Randomized Neural Networks
- Solving Tiling Enumeration Problems by Tensor Network Contractions
BibTeX
@misc{xu2026physelite,
title = {{PhysElite: How Far Are LLMs from Solving Olympiad-Level Physics Problems?}},
author = {Ruoran Xu and Wending Gao and Liyunfeng Chen and Aixin Shi and Haoyu Cheng and Zixiang Fang and Yiqiang Zou and Qiufeng Wang},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2608.25097},
eprint = {2608.25097},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.25097},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - PhysElite: How Far Are LLMs from Solving Olympiad-Level Physics Problems?
AU - Ruoran Xu
AU - Wending Gao
AU - Liyunfeng Chen
AU - Aixin Shi
AU - Haoyu Cheng
AU - Zixiang Fang
AU - Yiqiang Zou
AU - Qiufeng Wang
PY - 2026
DA - 2026/10/12
PB - arXiv
DO - 10.48550/arxiv.2608.25097
UR - https://arxiv.org/abs/2608.25097
AB - Understanding how (multimodal) large language models perform on physics problems requires benchmarks that reflect the difficulty and breadth of expert-level physical reasoning. Existing physics benchmarks remain limited in the following two important ways: (1) short of high-difficulty datasets, and (2) lack of comprehensive coverage of visual forms, knowledge points, and step-by-step solution processes. As a result, model performance on current datasets may not be fully representative of their ability to solve complex physics problems. To address these issues, we present PhysElite, a large-scale bilingual multimodal benchmark for Olympiad-level physics reasoning. PhysElite contains 11,586 Olympiad-tier problems. For each problem, we provide corresponding visual diagrams, step-by-step bilingual Chinese-English solution derivations, and the final answer. We benchmark 18 open-source and closed-source MLLMs, and find that even the strongest model reaches only 33.7% answer accuracy. We additionally conduct step-level process evaluation to diagnose where models fail in the reasoning chain. Our datasets are released at https://huggingface.co/datasets/physelite/PhysElite.
ER -