Training a Scientific Reasoning Model for Chemistry

TL;DR

This paper introduces ether0, a large language model trained for scientific reasoning in chemistry, capable of understanding and generating chemical structures and solving diverse chemistry problems with high accuracy and data efficiency.

Contribution

The authors demonstrate that reasoning models can be effectively adapted to chemistry without domain-specific pretraining, using reinforcement learning on a large, diverse dataset.

Findings

ether0 outperforms existing chemistry models and human experts on molecular design tasks.

The model requires significantly less data than specialized models.

It can reason in natural language and generate chemical structures.

Abstract

Reasoning models are large language models that emit a long chain-of-thought before answering, providing both higher accuracy and explicit reasoning for their response. A major question has been whether language model reasoning generalizes beyond mathematics, programming, and logic, where most previous work has focused. We demonstrate that reasoning models can be post-trained for chemistry without additional domain pretraining, and require substantially less data compared to contemporary domain-specific models. We report ether0, a 24B parameter LLM (based on Mistral-Small-24B) that can reason in natural language and respond with chemical structures. This reasoning model was trained with reinforcement learning on 640,730 experimentally-grounded chemistry problems across 375 tasks ranging from synthesizability, to blood-brain barrier permeability, to human receptor activity, to scent. Our model exceeds general-purpose chemistry models, frontier models, and human experts on molecular design tasks. It is also more data efficient relative to specialized models. We anticipate that this method can be applied to train data-efficient language models specialized for tasks across a wide variety of scientific domains.