SureRoute: Toward a Hallucination-Free Self-Improving Platform for Retrosynthesis
Abstract
AI models, including large language models, are increasingly integrated into scientific discovery workflows, yet they remain prone to hallucination. In experimental sciences, such errors translate directly into failed wet-lab validations and wasted resources; in self-improving agentic systems, confident errors risk being reinforced rather than corrected. Retrosynthesis provides a representative example of this failure mode: existing models can generate chemically plausible routes, but cannot reliably determine which routes are experimentally feasible. We define \textbf{Chemical Hallucination} as a route that appears valid yet fails under competing reactive sites, unresolved selectivity, or missing mechanistic support, a failure largely invisible to the Recall@K metric. We introduce \textbf{SureRoute}, a chemical verifier-anchored retrosynthesis platform that suppresses Chemical Hallucination. SureRoute combines a multi-model ensemble, data asset retrieval, and \textbf{ChemHarness}, an executable chemical intuition engine for route verification and reliability-first ranking. On a benchmark of 350 real-world industrial targets, SureRoute reaches 74.3\% recall@1, 2.2--3.5× that of seven single-step models and three frontier LLMs, while cutting top-1 Chemical Hallucination to 4.6\%, a 4--6× reduction relative to frontier LLMs. As a model-agnostic reranker, ChemHarness drives detectable hallucination toward near-zero across arbitrary backbone candidates. SureRoute shows that reliable scientific AI requires not only strong generation, but executable verification.
Related
- RetroGEF: Dynamic Graph Edit Flow for Single-Step Retrosynthesis
- An integration-free CRISPR platform for genetic dissection of respiratory virus host dependencies in human nasal epithelium
- SynBioCrow: provenance-preserving integration of complementary retrosynthesis engines for biosynthetic pathway design
- Projected Gradient Descent for Hallucination-Based Protein Binder Design
- Improving Molecular Docking Pose Recovery through Targeted Side-Chain Monte Carlo Sampling of an Apo Structure
BibTeX
@misc{zhou2026sureroute,
title = {{SureRoute: Toward a Hallucination-Free Self-Improving Platform for Retrosynthesis}},
author = {Jieli Zhou and Naiwu Chen and Longzhang Liu and Peiyu Zhang},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.05450},
eprint = {2609.05450},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.05450},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - SureRoute: Toward a Hallucination-Free Self-Improving Platform for Retrosynthesis
AU - Jieli Zhou
AU - Naiwu Chen
AU - Longzhang Liu
AU - Peiyu Zhang
PY - 2026
DA - 2026/09/01
PB - arXiv
DO - 10.48550/arxiv.2609.05450
UR - https://arxiv.org/abs/2609.05450
AB - AI models, including large language models, are increasingly integrated into scientific discovery workflows, yet they remain prone to hallucination. In experimental sciences, such errors translate directly into failed wet-lab validations and wasted resources; in self-improving agentic systems, confident errors risk being reinforced rather than corrected. Retrosynthesis provides a representative example of this failure mode: existing models can generate chemically plausible routes, but cannot reliably determine which routes are experimentally feasible. We define \textbf{Chemical Hallucination} as a route that appears valid yet fails under competing reactive sites, unresolved selectivity, or missing mechanistic support, a failure largely invisible to the Recall@K metric. We introduce \textbf{SureRoute}, a chemical verifier-anchored retrosynthesis platform that suppresses Chemical Hallucination. SureRoute combines a multi-model ensemble, data asset retrieval, and \textbf{ChemHarness}, an executable chemical intuition engine for route verification and reliability-first ranking. On a benchmark of 350 real-world industrial targets, SureRoute reaches 74.3\% recall@1, 2.2--3.5× that of seven single-step models and three frontier LLMs, while cutting top-1 Chemical Hallucination to 4.6\%, a 4--6× reduction relative to frontier LLMs. As a model-agnostic reranker, ChemHarness drives detectable hallucination toward near-zero across arbitrary backbone candidates. SureRoute shows that reliable scientific AI requires not only strong generation, but executable verification.
ER -