ScentGen: Hierarchical Multimodal Olfactory Semantic Modeling for Molecular Odor Description Generation

2026-09-26

AI Summary

This study addresses the limited expressiveness and poor interpretability of conventional discrete odor labels by proposing a hierarchical multimodal olfactory semantic modeling framework. Methodologically, it introduces the first end-to-end generative paradigm that translates molecular structures into natural language odor descriptions. Specifically, the approach integrates SMILES sequences, molecular graphs, and 3D conformations for multimodal encoding, mapping molecular representations into the latent space of large language models via continuous prompt generation. As a key contribution, a corresponding multimodal paired dataset is constructed to support this framework. Experimental results demonstrate that the proposed method generates coherent and highly expressive odor descriptions, significantly outperforming traditional discrete label prediction approaches. Ultimately, this work enhances both the flexibility and interpretability of molecular odor understanding.

Abstract

In this paper, we introduce a molecular odor description generation task, which aims to generate natural language odor descriptions from molecular structures. Unlike conventional methods that describe molecular odor using discrete labels, this task generates expressive and human-interpretable sensory descriptions. To address this task, we propose a hierarchical multimodal olfactory semantic modeling framework, named ScentGen. ScentGen consists of three key components: an odor semantic planner, a semantic adapter, and a description generator. The odor semantic planner integrates complementary molecular information from 1D SMILES sequences, 2D molecular graphs, and 3D molecular conformations to learn discriminative and structured olfactory semantics. The semantic adapter further maps the learned olfactory representation into the hidden space of a large language model, transforming molecular odor semantics into language-compatible continuous prompts.Conditioned on these prompts, the description generator produces coherent odor descriptions that reflect plausible sensory characteristics of the input molecule. Considering the lack of molecular datasets with natural language odor descriptions, we further construct a molecular odor description dataset containing paired multimodal molecular representations and human-interpretable odor descriptions. Extensive experiments demonstrate that ScentGen generates coherent and expressive odor descriptions, providing a more flexible solution for molecular odor understanding beyond discrete odor label prediction.

Problem

• molecular odor description generation
• olfactory semantic modeling
• multimodal molecular representation
• natural language odor description

Innovation

• Molecular Odor Description Generation
• Hierarchical Multimodal Modeling
• Large Language Model
• Olfactory Semantics
• Semantic Adapter