RetroReasoner: A Reasoning LLM for Strategic Retrosynthesis Prediction
cs.LG, cs.AI
Submitted: 2026-03-13
Updated: 2026-09-01
Comments: 36 pages, 20 figures
Code: https://github.com/KU-AGI/RetroReasoner
License: http://creativecommons.org/licenses/by-nc-sa/4.0/
The gist: Retrosynthesis prediction aims to identify reactants that can synthesize a given product molecule.
Terminology
Abstract
Retrosynthesis prediction aims to identify reactants that can synthesize a given product molecule. Although molecular large language models (LLMs) have recently shown promising results, most existing methods either generate reactants directly or provide only generic product-level analysis, without explicitly reasoning about bond-disconnection strategies that justify specific reactant choices. This paper proposes RetroReasoner, a retrosynthetic reasoning model that captures chemists' strategic disconnection-based thinking. RetroReasoner is trained with supervised fine-tuning and reinforcement learning. For supervised fine-tuning, SyntheticRetro generates structured disconnection rationales paired with reactant predictions. For reinforcement learning, a round-trip reward evaluates predicted reactants by passing them through a forward synthesis model and rewarding predictions that reconstruct the original product. RetroReasoner can also be applied to multi-step retrosynthetic planning by incorporating it into a parallelized Monte Carlo tree search framework, reducing search time while increasing the number and diversity of valid synthetic pathways. Experimental results show that RetroReasoner outperforms prior baselines, including not only molecular LLMs but also retrosynthesis-specific expert models, and generates a broader range of feasible reactant proposals, especially for challenging reaction instances. The code is available at https://github.com/KU-AGI/RetroReasoner.
Sources
- Mol-Instructions: A Large-Scale Biomolecular Instruction Dataset for Large Language Models
- gpt-oss-120b & gpt-oss-20b Model Card
- PRESTO: Progressive Pretraining Enhances Synthetic Chemistry Outcomes
- Mol-LLM: Multimodal Generalist Molecular LLM with Improved Graph Utilization
- Modern Hopfield Networks for Few- and Zero-Shot Reaction Template Prediction
- Retro-Expert: Collaborative Reasoning for Interpretable Retrosynthesis
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Watch the Unobserved: A Simple Approach to Parallelizing Monte Carlo Tree Search
- Decoupled Weight Decay Regularization
- HybridFlow: A Flexible and Efficient RLHF Framework
- Training a Scientific Reasoning Model for Chemistry
- OpenAI GPT-5 System Card
- BioT5+: Towards Generalized Biological Understanding with IUPAC Integration and Multi-task Tuning
- AOT*: Efficient Synthesis Planning via LLM-Empowered AND-OR Tree Search
- Proximal Policy Optimization Algorithms
- Chem-R: Learning to Reason as a Chemist
- ChemDFM-R: A Chemical Reasoning LLM Enhanced with Atomized Chemical Knowledge
- TempRe: Template generation for single and direct multi-step retrosynthesis
- Qwen3 Technical Report
- LlaSMol: Advancing Large Language Models for Chemistry with a Large-Scale, Comprehensive, High-Quality Instruction Tuning Dataset
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks