Rachel: A general-purpose language model directs and revises retrosynthetic routes
Abstract
Retrosynthetic planning advances through decisions that reshape the remaining chemical problem: a locally plausible disconnection can leave precursors whose chemoselectivity constraints complicate the rest of the route. Existing planners often channel model proposals through search or template procedures, leaving open whether a general-purpose large language model (LLM) can itself sustain and revise route strategy. We developed Rachel, a stateful environment that executes and checks LLM-directed chemistry but prescribes neither a search policy nor a stopping rule. Without supplied reference routes or route-level solutions, GPT-5.5 achieved strict closure for 111 of 120 PaRoutes120 targets and 24 of 25 targets in the separate RF25 difficult-target cohort. RF25 was drawn largely from studies published after GPT-5.5's reported knowledge cutoff. Closure required complete routes and independent source resolution of every terminal precursor after planning. On a shared PaRoutes subset, forward-model support exceeded that of most comparator methods, and Rachel received the highest mean overall route score from both method-blinded LLM evaluators. Recorded trajectories showed continued model-proposed chemistry, with revised strategies carried into subsequent steps. Replacing LLM route decisions with fixed policies reduced strict closure to 6-15/120 despite continued local chemical execution; restricting planning support also reduced closure in RF25. Within Rachel, a general-purpose LLM coordinated successive chemical choices and revised its strategy as earlier decisions reshaped the remaining problems.
Related
- VIBE-QC: an open-source, general-purpose electronic-structure code for molecules and solids developed with generative AI
- DeepH-pack: a general-purpose neural network package for deep-learning electronic structure calculations
- General-purpose foundation models surpass specialized ones on chemical property prediction tasks
- From Retrosynthetic Prediction to the Bench: Experimental Evaluation of Computer-Generated Synthetic Routes
- Near-Complete Synthon Library via Retrosynthetic Analysis for Automated Organic Synthesis
BibTeX
@misc{li2026rachel,
title = {{Rachel: A general-purpose language model directs and revises retrosynthetic routes}},
author = {Qisheng Li and Shunchao Jiang and Chen Qi and Xin Su and Da Han and Guangyong Chen},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.25118},
eprint = {2609.25118},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.25118},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Rachel: A general-purpose language model directs and revises retrosynthetic routes
AU - Qisheng Li
AU - Shunchao Jiang
AU - Chen Qi
AU - Xin Su
AU - Da Han
AU - Guangyong Chen
PY - 2026
DA - 2026/09/20
PB - arXiv
DO - 10.48550/arxiv.2609.25118
UR - https://arxiv.org/abs/2609.25118
AB - Retrosynthetic planning advances through decisions that reshape the remaining chemical problem: a locally plausible disconnection can leave precursors whose chemoselectivity constraints complicate the rest of the route. Existing planners often channel model proposals through search or template procedures, leaving open whether a general-purpose large language model (LLM) can itself sustain and revise route strategy. We developed Rachel, a stateful environment that executes and checks LLM-directed chemistry but prescribes neither a search policy nor a stopping rule. Without supplied reference routes or route-level solutions, GPT-5.5 achieved strict closure for 111 of 120 PaRoutes120 targets and 24 of 25 targets in the separate RF25 difficult-target cohort. RF25 was drawn largely from studies published after GPT-5.5's reported knowledge cutoff. Closure required complete routes and independent source resolution of every terminal precursor after planning. On a shared PaRoutes subset, forward-model support exceeded that of most comparator methods, and Rachel received the highest mean overall route score from both method-blinded LLM evaluators. Recorded trajectories showed continued model-proposed chemistry, with revised strategies carried into subsequent steps. Replacing LLM route decisions with fixed policies reduced strict closure to 6-15/120 despite continued local chemical execution; restricting planning support also reduced closure in RF25. Within Rachel, a general-purpose LLM coordinated successive chemical choices and revised its strategy as earlier decisions reshaped the remaining problems.
ER -