Can Transformers Smell Like Humans?
This addresses the under-explored challenge of olfactory perception in AI, providing a method to bridge chemical data and human sensory experience, though it is incremental as it applies existing transformer techniques to a new domain.
The paper tackled the problem of aligning machine learning models with human olfactory perception by showing that pre-trained transformer models on chemical structures can predict human odor labels, ratings, and similarities with high alignment, achieving strong correlation scores (e.g., up to 0.85 on specific datasets).
The human brain encodes stimuli from the environment into representations that form a sensory perception of the world. Despite recent advances in understanding visual and auditory perception, olfactory perception remains an under-explored topic in the machine learning community due to the lack of large-scale datasets annotated with labels of human olfactory perception. In this work, we ask the question of whether pre-trained transformer models of chemical structures encode representations that are aligned with human olfactory perception, i.e., can transformers smell like humans? We demonstrate that representations encoded from transformers pre-trained on general chemical structures are highly aligned with human olfactory perception. We use multiple datasets and different types of perceptual representations to show that the representations encoded by transformer models are able to predict: (i) labels associated with odorants provided by experts; (ii) continuous ratings provided by human participants with respect to pre-defined descriptors; and (iii) similarity ratings between odorants provided by human participants. Finally, we evaluate the extent to which this alignment is associated with physicochemical features of odorants known to be relevant for olfactory decoding.