{"slug": "how-good-are-frontier-models-at-physics", "title": "How good are frontier models at physics?", "summary": "An expert re-grading audit of six widely used physics benchmarks found that most cases scored as incorrect reflected grader errors, flawed reference solutions, or ambiguous questions rather than genuine model failures, according to an arXiv paper submitted on 11 September 2026. After experts corrected reference solutions and repaired or excluded flawed questions, GPT-5.6-Sol's measured mean@4 rose from 47.3% to 78.7% on HLE-Physics and from 61.0% to 87.2% on CMT-Benchmark, while its corrected pass@4 reached 94.4% on the 54 retained CritPt challenges. Corrected scores also rose substantially on the audited subsets of UGPhysics, PRISM-Physics, and PHYBench, which the authors say indicates current benchmarks substantially understate frontier models' ability to solve well-posed physics problems and are nearing saturation on closed-ended tasks.", "body_md": "# Computer Science > Artificial Intelligence\n\n  [Submitted on 11 Sep 2026]\n\n# Title:How Good Are Frontier Models at Physics? Expert Re-Grading Reveals Broken Evaluations and Near-Saturation of Leading Benchmarks\n\n[View PDF](https://arxiv.org/pdf/2609.13009)\n\n[HTML (experimental)](https://arxiv.org/html/2609.13009v1)\n\nAbstract:Low reported scores on leading physics benchmarks, including those featured in the Artificial Analysis Intelligence Index (2026), suggest that frontier language models still struggle with advanced physics, a demanding test of their scientific reasoning and quantitative problem-solving abilities. Yet this impression does not always align with domain experts' experiences using these models in their work. We revisit these reported findings by evaluating frontier models on six widely used physics benchmarks and auditing them with experts, focusing on text-only problems with verifiable final answers. For each subfield of physics, faculty and graduate researchers with relevant expertise carefully review problem statements, reference solutions, and model responses to distinguish genuine model errors from grader errors, incorrect reference solutions, and ambiguous or underspecified questions. Most audited cases initially evaluated as incorrect reflect these benchmarking issues rather than errors in the models' physics reasoning. We then ask experts to address these benchmarking issues by correcting erroneous reference solutions and repairing or excluding flawed questions. We find that GPT-5.6-Sol's measured mean@4 rises from 47.3% to 78.7% on HLE-Physics and from 61.0% to 87.2% on CMT-Benchmark, while its corrected pass@4 reaches 94.4% on the 54 retained CritPt challenges. Corrected scores are computed on the retained evaluation subsets following expert review. Scores on the audited subsets of UGPhysics, PRISM-Physics, and PHYBench also rise substantially after correction. These findings suggest that current benchmarks substantially understate frontier models' ability to solve well-posed physics problems. Near-saturation on these closed-ended tasks highlights the need for more demanding, expert-validated evaluations.\n    \n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer \n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers \n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps \n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations \n\n*(*[What are Smart Citations?](https://www.scite.ai/))\n# Code, Data and Media Associated with this Article\n\nalphaXiv \n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers \n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub \n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub \n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face \n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast \n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))\n# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower \n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender \n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))\n# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.\n\nHave an idea for a project that will add value for arXiv's community? [**Learn more about arXivLabs**](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/how-good-are-frontier-models-at-physics", "canonical_source": "https://arxiv.org/abs/2609.13009", "published_at": "2026-09-16 19:19:08+00:00", "updated_at": "2026-09-16 19:42:09.263054+00:00", "lang": "en", "topics": ["ai-research", "large-language-models", "artificial-intelligence", "ai-safety"], "entities": ["GPT-5.6-Sol", "HLE-Physics", "CMT-Benchmark", "CritPt", "UGPhysics", "PRISM-Physics", "PHYBench", "Artificial Analysis Intelligence Index"], "alternates": {"html": "https://wpnews.pro/news/how-good-are-frontier-models-at-physics", "markdown": "https://wpnews.pro/news/how-good-are-frontier-models-at-physics.md", "text": "https://wpnews.pro/news/how-good-are-frontier-models-at-physics.txt", "jsonld": "https://wpnews.pro/news/how-good-are-frontier-models-at-physics.jsonld"}}