AI & ML Collision

We can now mathematically map the chemical structure of a molecule directly to the human linguistic experience of its smell.

April 14, 2026

Original Paper

NOSE: Neural Olfactory-Semantic Embedding with Tri-Modal Orthogonal Contrastive Learning

Yanyi Su, Hongshuai Wang, Zhifeng Gao, Jun Cheng

arXiv · 2604.10452

The Takeaway

Using tri-modal orthogonal contrastive learning, this system aligns molecular structure, receptors, and natural language descriptions. It enables a direct algebraic bridge between chemical coordinates and subjective human sensory experience.

From the abstract

Olfaction lies at the intersection of chemical structure, neural encoding, and linguistic perception, yet existing representation methods fail to fully capture this pathway. Current approaches typically model only isolated segments of the olfactory pathway, overlooking the complete chain from molecule to receptors to linguistic descriptions. Such fragmentation yields learned embeddings that lack both biological grounding and semantic interpretability. We propose NOSE (Neural Olfactory-Semantic E