探究表格基础模型是否掌握物理规律，发现其无法理解单位与确定性机制。
Do Tabular Foundation Models Know Physics? Contamination, Units, and the Deterministic Limit
- 直接测试四种表格模型对316个物理方程数据的泛化能力。
- 模型在域内域外均表现优异，但无法正确处理物理单位和噪声消失情形。
- 适合关注模型物理先验、可解释性的研究人员参考。
表格基础模型（TFMs）以填补表格的方式学习，而物理测量数据多以表格形式呈现。它们是否在学习过程中掌握了物理知识？由于其贝叶斯构建特性，关键在于其先验包含什么。我们直接评估了四种模型（TabPFN-3、TabICLv2、TabDPT 和 Real-TabPFN-2.5）在316个物理方程生成的数据集上的表现，涵盖域内与域外样本，并与六种基线对比。结果表明，这些模型在未微调时即表现出色，经微调后更优。然而，我们揭示其先验无法表示无噪声机制或物理单位，因此虽能插值物理规律，尚不能作为真正的物理模型使用。
原文摘要 · Abstract (English)
Tabular foundation models (TFMs) learn to fill in tables the way language models fill in text, and tables are arguably the format in which most physical measurement arrives. Did they learn any physics in the process? They are Bayesian by construction, so the question is what their prior contains. We probe it directly, evaluating four of them (TabPFN-3, TabICLv2, TabDPT and Real-TabPFN-2.5) against six baselines on datasets sampled from 316 physical equations, in and out of domain. TFMs dominate, out of the box and after tuning. But we show that their prior can represent neither a noiseless mechanism nor physical units, which is why they interpolate physics without yet being able to act as physical models.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。