CLEANMOL, a novel framework, enhances structural comprehension in large language models for molecular science by formulating SMILES parsing into structured tasks, improving performance on Mol-Instructions.
Large language models (LLMs) are increasingly recognized as powerful tools
for scientific discovery, particularly in molecular science. A fundamental
requirement for these models is the ability to accurately understand molecular
structures, commonly encoded in the SMILES representation. However, current
LLMs struggle to interpret SMILES, even failing to carry out basic tasks such
as counting molecular rings. To address this limitation, we introduce CLEANMOL,
a novel framework that formulates SMILES parsing into a suite of clean and
deterministic tasks explicitly designed to promote graph-level molecular
comprehension. These tasks span from subgraph matching to global graph
matching, providing structured supervision aligned with molecular structural
properties. We construct a molecular pretraining dataset with adaptive
difficulty scoring and pre-train open-source LLMs on these tasks. Our results
show that CLEANMOL not only enhances structural comprehension but also achieves
the best or competes with the baseline on the Mol-Instructions benchmark.