NeFut Logo NeFut
Admin Login

[CS.AI] Revolutionary Advances in Molecular Property Prediction with Small Language Models

Published at: 2026-07-16 22:00 Last updated: 2026-07-17 08:45
#AI #Machine Learning #Graph

Abstract

Small language models (SLMs) have shown promise for zero-shot molecular property prediction from SMILES strings, yet they often suffer from structural blindness because sequence representations under-specify key graph-topological cues. We propose a modular Context-Augmented Prompting framework that enables agentic tool use at inference time: a trained GNN expert model provides a predictive hint with confidence, and a GNN extracts an instance-specific explanatory subgraph (e.g., a subgraph SMILES and an accompanying explanatory paragraph).

We evaluate three commonly used SLMs on MUTAG and Tox21 under five prompting configurations ranging from SMILES-only to using all available tools at hand. Across two datasets, enriching prompts with graph-derived context yields substantial accuracy gains, often exceeding 25% relative improvement and up to 74% on Tox21. We further validate the functional relevance of the extracted motifs via a necessity-based edge-drop intervention. Despite the observed gains, a persistent gap remains to specialized GNN models, highlighting both the value and limits of text-conditioned reasoning for molecular structure.

Blogger's Review: This research demonstrates the potential of combining small language models with graph neural networks, particularly in the field of molecular property prediction. The significant improvement in accuracy through context-augmented prompting offers new insights for future molecular modeling and cheminformatics.

Original Source: https://arxiv.org/abs/2607.13115

[h] Back to Home