Foundation Molecular Grammar: Multi-Modal Foundation Models Induce Interpretable Molecular Graph Languages
Michael Sun, Weize Yuan, Gang Liu, Wojciech Matusik, Jie Chen · 2025
Read Foundation Molecular Grammar: Multi-Modal Foundation Models Induce Interpretable Molecular Graph Languages by Michael Sun, Weize Yuan, Gang Liu, Wojciech Matusik, Jie Chen (2025) 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-2505-22948.txt.
Introduction
Michael Sun and co-authors propose, in a 2025 paper, a way to teach generative chemistry models a readable grammar of molecules. Graph grammars can make molecule generation more interpretable and data efficient. But learning such a grammar has relied on expert labels or shaky rules of thumb, which limits trust.
Their method, Foundation Molecular Grammar, leans on large multimodal foundation models and their stored chemical knowledge. It renders each molecule as an image, describes it in text, and aligns the two views through prompt based learning. The induced grammar can drop into existing pipelines for molecule generation and property prediction. In tests, it did well on whether molecules can be made, how varied they are, and how little data it needs. Its chemical rules stay readable by humans, which suits automated discovery work.
Catalog record
| CREATOR | Michael Sun, Weize Yuan, Gang Liu, Wojciech Matusik, Jie Chen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 177 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-22948