Tag
The paper introduces Verification-Guided Reflow (VGR), which integrates reaction verification as a decision action in AND/OR search for retrosynthesis planning, proposing early pruning and delayed guidance to boost verified solvability from 51.1% to 71.1% on Retro190.
Researchers introduce Solvent, an 8B LLM that extends long-horizon reasoning to multi-step retrosynthesis through a two-round post-training curriculum combining RL and iterative route revision, achieving 61.86% across four benchmarks and rivaling frontier trillion-parameter models.
RetroGEF is a flow-based generative model for single-step retrosynthesis that models molecular transformations and variable graph size within one unified process, achieving state-of-the-art results on standard retrosynthesis benchmarks. It constructs reactant graphs directly from product–reactant pairs without requiring a fixed graph canvas or a prescribed edit order.
Microsoft Research's RetroChimera, a predictive model for retrosynthesis, is published in Nature, helping to accelerate chemical synthesis for advances in medicine, materials, and agriculture.
Synthon 350M, a single-step retrosynthesis model, is launched by rasyn_lab, achieving state-of-the-art performance on the USPTO-50K benchmark with top-1 accuracy of 66.23% in single-shot settings.
This paper introduces an enhanced C3LM model for single-step retrosynthesis using Top-K prompting and a larger training dataset, demonstrating competitive performance with conventional models on the URSA-expert-2026 benchmark.
Ariadne is a decoder-only route generator for retrosynthetic planning that frames the target, optional constraints, and route as a prompt-completion sequence, achieving superior performance with much less computation compared to traditional search-based planners.
This paper presents ConRetroBert, a dual encoder framework for template-based single-step retrosynthesis that uses contrastive pretraining and listwise ranking to improve template prediction accuracy, achieving up to 75.4% top-1 accuracy on the USPTO-50k benchmark while maintaining interpretability.
EPFL researchers developed Synthegy, an AI framework that uses large language models to guide chemical retrosynthesis and reaction mechanism analysis through natural language instructions, significantly improving strategic planning for chemists.