Reasoning-Driven Retrosynthesis Prediction with Large Language Models via Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Situo, Li, Hanqi, Chen, Lu, Zhao, Zihan, Lin, Xuanze, Zhu, Zichen, Chen, Bo, Chen, Xin, Yu, Kai
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908462269595648
author Zhang, Situo
Li, Hanqi
Chen, Lu
Zhao, Zihan
Lin, Xuanze
Zhu, Zichen
Chen, Bo
Chen, Xin
Yu, Kai
author_facet Zhang, Situo
Li, Hanqi
Chen, Lu
Zhao, Zihan
Lin, Xuanze
Zhu, Zichen
Chen, Bo
Chen, Xin
Yu, Kai
contents Retrosynthesis planning, essential in organic synthesis and drug discovery, has greatly benefited from recent AI-driven advancements. Nevertheless, existing methods frequently face limitations in both applicability and explainability. Traditional graph-based and sequence-to-sequence models often lack generalized chemical knowledge, leading to predictions that are neither consistently accurate nor easily explainable. To address these challenges, we introduce RetroDFM-R, a reasoning-based large language model (LLM) designed specifically for chemical retrosynthesis. Leveraging large-scale reinforcement learning guided by chemically verifiable rewards, RetroDFM-R significantly enhances prediction accuracy and explainability. Comprehensive evaluations demonstrate that RetroDFM-R significantly outperforms state-of-the-art methods, achieving a top-1 accuracy of 65.0% on the USPTO-50K benchmark. Double-blind human assessments further validate the chemical plausibility and practical utility of RetroDFM-R's predictions. RetroDFM-R also accurately predicts multistep retrosynthetic routes reported in the literature for both real-world drug molecules and perovskite materials. Crucially, the model's explicit reasoning process provides human-interpretable insights, thereby enhancing trust and practical value in real-world retrosynthesis applications.
format Preprint
id arxiv_https___arxiv_org_abs_2507_17448
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Reasoning-Driven Retrosynthesis Prediction with Large Language Models via Reinforcement Learning
Zhang, Situo
Li, Hanqi
Chen, Lu
Zhao, Zihan
Lin, Xuanze
Zhu, Zichen
Chen, Bo
Chen, Xin
Yu, Kai
Computational Engineering, Finance, and Science
Artificial Intelligence
Chemical Physics
Retrosynthesis planning, essential in organic synthesis and drug discovery, has greatly benefited from recent AI-driven advancements. Nevertheless, existing methods frequently face limitations in both applicability and explainability. Traditional graph-based and sequence-to-sequence models often lack generalized chemical knowledge, leading to predictions that are neither consistently accurate nor easily explainable. To address these challenges, we introduce RetroDFM-R, a reasoning-based large language model (LLM) designed specifically for chemical retrosynthesis. Leveraging large-scale reinforcement learning guided by chemically verifiable rewards, RetroDFM-R significantly enhances prediction accuracy and explainability. Comprehensive evaluations demonstrate that RetroDFM-R significantly outperforms state-of-the-art methods, achieving a top-1 accuracy of 65.0% on the USPTO-50K benchmark. Double-blind human assessments further validate the chemical plausibility and practical utility of RetroDFM-R's predictions. RetroDFM-R also accurately predicts multistep retrosynthetic routes reported in the literature for both real-world drug molecules and perovskite materials. Crucially, the model's explicit reasoning process provides human-interpretable insights, thereby enhancing trust and practical value in real-world retrosynthesis applications.
title Reasoning-Driven Retrosynthesis Prediction with Large Language Models via Reinforcement Learning
topic Computational Engineering, Finance, and Science
Artificial Intelligence
Chemical Physics
url https://arxiv.org/abs/2507.17448