Physics as the Inductive Bias for Causal Discovery
Jianhong Chen, Naichen Shi, Xubo Yue · 2026
Read Physics as the Inductive Bias for Causal Discovery by Jianhong Chen, Naichen Shi, Xubo Yue (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-04907.txt.
Description
Causal discovery means working out which parts of a system drive which others, using data alone, and this long work of about 100 pages mixes it with physics. The paper comes from Jianhong Chen and 2 co-authors. Pure data methods struggle with systems that loop back on themselves or change over time, which is common in the real world.
Their framework mixes in partial physics. Known dynamics go into the drift part of a stochastic differential equation, while the random part is free to capture unknown links. A new estimator with a stabilizing trick keeps the math well behaved, and the authors prove recovery guarantees for both stable and unstable systems. Tests on predator-prey and chaotic benchmarks, plus real epidemic data, beat purely data-driven rivals.
Catalog record
| CREATOR | Jianhong Chen, Naichen Shi, Xubo Yue |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 122 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-04907