AROMMA Agent

Code: https://github.com/DGIST-Distributed-AI-Lab/aromma-agent

AROMMA Agent is an LLM-powered odor prediction and recommendation system built on top of AROMMA (ICASSP 2026).

It supports the following capabilities:

• Odor prediction from a molecule or a binary mixture given SMILES strings
• Molecule recommendation for composing a target scent described in natural language

Olfaction to Language: odor prediction

When a user requests the scent of a specific molecule or mixture, the trained AROMMA predicts odor descriptors, and the large language model transforms them into natural-language sentences.

Language to Olfaction: molecule recommendation

When a user enters a description of a desired scent, the large language model selects suitable labels from 152 odor descriptors. It then searches the database to identify and present molecules that produce that scent.