Evaluating the Progression of Large Language Model Capabilities for Small-Molecule Drug Design

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chennakesavalu, Shriram, Shmilovich, Kirill, Weir, Hayley, Grambow, Colin, Bradshaw, John, Suriana, Patricia, Cheng, Chen, Chuang, Kangway
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917416350515200
author Chennakesavalu, Shriram
Shmilovich, Kirill
Weir, Hayley
Grambow, Colin
Bradshaw, John
Suriana, Patricia
Cheng, Chen
Chuang, Kangway
author_facet Chennakesavalu, Shriram
Shmilovich, Kirill
Weir, Hayley
Grambow, Colin
Bradshaw, John
Suriana, Patricia
Cheng, Chen
Chuang, Kangway
contents Large Language Models (LLMs) have the potential to accelerate small molecule drug design due to their ability to reason about information from diverse sources and formats. However, their practical utility remains unclear due to the lack of benchmarks that reflect real-world scenarios. In this work, we introduce a suite of chemically-grounded tasks spanning molecular property prediction, molecular representation transformations, and molecular design. Importantly, we formulate these tasks as reinforcement learning (RL) environments, enabling a unified approach for evaluation and post-training. Across three model families, we find that frontier models are increasingly proficient at chemical tasks, but that there is significant room for improvement, especially in experimental settings with low data. Critically, we show that RL-based post-training can substantially improve performance. A smaller model post-trained on our environments becomes competitive with state-of-the-art frontier models, despite a significantly weaker base model. This suggests a practical route toward employing LLMs in drug discovery; by combining carefully-designed evaluation tasks with targeted post-training, we can both elucidate and close critical capability gaps.
format Preprint
id arxiv_https___arxiv_org_abs_2604_16279
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Evaluating the Progression of Large Language Model Capabilities for Small-Molecule Drug Design
Chennakesavalu, Shriram
Shmilovich, Kirill
Weir, Hayley
Grambow, Colin
Bradshaw, John
Suriana, Patricia
Cheng, Chen
Chuang, Kangway
Machine Learning
Chemical Physics
Large Language Models (LLMs) have the potential to accelerate small molecule drug design due to their ability to reason about information from diverse sources and formats. However, their practical utility remains unclear due to the lack of benchmarks that reflect real-world scenarios. In this work, we introduce a suite of chemically-grounded tasks spanning molecular property prediction, molecular representation transformations, and molecular design. Importantly, we formulate these tasks as reinforcement learning (RL) environments, enabling a unified approach for evaluation and post-training. Across three model families, we find that frontier models are increasingly proficient at chemical tasks, but that there is significant room for improvement, especially in experimental settings with low data. Critically, we show that RL-based post-training can substantially improve performance. A smaller model post-trained on our environments becomes competitive with state-of-the-art frontier models, despite a significantly weaker base model. This suggests a practical route toward employing LLMs in drug discovery; by combining carefully-designed evaluation tasks with targeted post-training, we can both elucidate and close critical capability gaps.
title Evaluating the Progression of Large Language Model Capabilities for Small-Molecule Drug Design
topic Machine Learning
Chemical Physics
url https://arxiv.org/abs/2604.16279