Language models for molecular design have scaled to hundreds of millions of parameters, yet how they learn chemical grammar is poorly understood. We train SMolLM, a 53K-parameter weight-shared transformer, to generate novel SMILES with 95% validity on the ZINC-250K drug-like-molecule benchmark, outperforming a standard GPT with 10 times more parameters. Mechanistically, the same block resolves SMILES constraints across passes in a fixed hierarchy: brackets first, rings second, and valence last, as shown by error classification and linear probing, with ablation isolating the bracket-matching head. Together, these results yield a compact, mechanistically interpretable molecular generator and a testbed for studying iterative computation in formal-language domains.
@article{arxiv.2605.06322,
title = {SMolLM: Small Language Models Learn Small Molecular Grammar},
author = {Akhil Jindal and Harang Ju},
journal= {arXiv preprint arXiv:2605.06322},
year = {2026}
}