Guardado en:
Detalles Bibliográficos
Autores principales: Zhang, Qingchuan, Cao, He, Li, Hao, Shao, Yanjun, Liu, Zhiyuan, Wang, Shihang, Xie, Shufang, Gao, Shenghua, Ye, Xinwu
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2605.10230
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916000211927040
author Zhang, Qingchuan
Cao, He
Li, Hao
Shao, Yanjun
Liu, Zhiyuan
Wang, Shihang
Xie, Shufang
Gao, Shenghua
Ye, Xinwu
author_facet Zhang, Qingchuan
Cao, He
Li, Hao
Shao, Yanjun
Liu, Zhiyuan
Wang, Shihang
Xie, Shufang
Gao, Shenghua
Ye, Xinwu
contents Molecular optimization seeks to improve a molecule through small structural edits while preserving similarity to the starting compound. Recent language-model approaches typically treat this task as prompt-conditioned sequence generation. However, relying on natural language introduces an inherent data-scaling bottleneck, often leads to chemical hallucinations, and ignores the strong context dependence of fragment effects. We present FORGE, a two-stage framework that reformulates molecular optimization as context-aware local editing. By utilizing automatically mined, verified low-to-high edit pairs instead of expensive human text annotations, Stage 1 ranks candidate fragments by their property contribution under the full molecular context to inject chemical prior, and Stage 2 generates explicit fragment replacements. Built on a compact 0.6B language model, FORGE further adapts to unseen black-box objectives through in-context demonstrations. Across Prompt-MolOpt, PMO-1k and ChemCoTBench, FORGE consistently outperforms prior methods, including substantially larger language models and graph methods. These results highlight the value of explicit fragment-level supervision as a more easily obtainable, scalable, and hallucination-less alternative to natural language training.
format Preprint
id arxiv_https___arxiv_org_abs_2605_10230
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle FORGE: Fragment-Oriented Ranking and Generation for Context-Aware Molecular Optimization
Zhang, Qingchuan
Cao, He
Li, Hao
Shao, Yanjun
Liu, Zhiyuan
Wang, Shihang
Xie, Shufang
Gao, Shenghua
Ye, Xinwu
Machine Learning
Molecular optimization seeks to improve a molecule through small structural edits while preserving similarity to the starting compound. Recent language-model approaches typically treat this task as prompt-conditioned sequence generation. However, relying on natural language introduces an inherent data-scaling bottleneck, often leads to chemical hallucinations, and ignores the strong context dependence of fragment effects. We present FORGE, a two-stage framework that reformulates molecular optimization as context-aware local editing. By utilizing automatically mined, verified low-to-high edit pairs instead of expensive human text annotations, Stage 1 ranks candidate fragments by their property contribution under the full molecular context to inject chemical prior, and Stage 2 generates explicit fragment replacements. Built on a compact 0.6B language model, FORGE further adapts to unseen black-box objectives through in-context demonstrations. Across Prompt-MolOpt, PMO-1k and ChemCoTBench, FORGE consistently outperforms prior methods, including substantially larger language models and graph methods. These results highlight the value of explicit fragment-level supervision as a more easily obtainable, scalable, and hallucination-less alternative to natural language training.
title FORGE: Fragment-Oriented Ranking and Generation for Context-Aware Molecular Optimization
topic Machine Learning
url https://arxiv.org/abs/2605.10230