Localize, Then Reason: Visual Latent Structural Reasoning for Molecular Properties and Edits

2026-08-13

AI Summary

Existing methods struggle to focus on chemically meaningful regions in molecular images prior to inference, hindering a deeper understanding of structure–property relationships. To address this limitation, this work proposes the Visual Latent Structure Reasoning (VLSR) framework, which introduces an end-to-end “locate-then-reason” cooperative mechanism. VLSR jointly optimizes region localization and property prediction within an implicit latent workspace through a unified deep learning architecture. This approach significantly enhances the model’s chemical awareness and reasoning efficiency, achieving a throughput 9.6 times higher than that of text-based baselines under identical experimental settings.

Abstract

Local chemical perception and property reasoning are both essential for understanding how molecular structure determines properties. Current LLM-based chemical reasoning methods either receive SMILES/molecular images together with descriptions of local motifs, or reason directly from molecular images. Neither approach enables the model to focus on chemically meaningful regions before reasoning. To address this gap, we propose Visual Latent Structural Reasoning (VLSR), an end-to-end framework that jointly learns localization and reasoning from molecular images. Central to our approach is a localize-then-reason strategy. VLSR first learns to locate chemically meaningful regions in a molecular image. It then reasons about their property effects in a compact latent workspace before producing the final answer. Under the same inference setup, this design achieves 9.6X higher throughput than a comparable textual-reasoning baseline.

Problem

• molecular properties
• visual reasoning
• localization
• chemical perception
• molecular images

Innovation

• Visual Latent Structural Reasoning
• localize-then-reason
• molecular image reasoning
• chemical perception
• latent workspace