FORGE: Fragment-Oriented Ranking and Generation for Context-Aware Molecular Optimization
Qingchuan Zhang, He Cao, Hao Li, Yanjun Shao, Zhiyuan Liu, Shihang Wang, Shufang Xie, Shenghua Gao, Xinwu Ye · 2026
Read FORGE: Fragment-Oriented Ranking and Generation for Context-Aware Molecular Optimization by Qingchuan Zhang, He Cao, Hao Li, Yanjun Shao, Zhiyuan Liu, Shihang Wang, Shufang Xie, Shenghua Gao, Xinwu Ye (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-10230.txt.
Description
Improving a drug-like molecule usually means small, careful edits, not a fresh design. Qingchuan Zhang and eight colleagues recast that task as context-aware local editing. Their 2026 paper presents FORGE, a two-stage framework for molecular optimization that decides where to edit a molecule and how to replace the chosen fragment.
Stage one ranks candidate fragments by their property contribution within the full molecular context, using automatically mined and verified edit pairs instead of costly human-written text. Stage two generates the explicit replacement. Built on a compact 0.6-billion-parameter language model, FORGE beat prior methods across three benchmarks, including much larger language models and graph methods. It also adapts to unseen black-box goals through in-context examples.
Catalog record
| CREATOR | Qingchuan Zhang, He Cao, Hao Li, Yanjun Shao, Zhiyuan Liu, Shihang Wang, Shufang Xie, Shenghua Gao, Xinwu Ye |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.10230