Chain-of-Generation: Progressive Latent Diffusion for Text-Guided Molecular Design
Lingxiao Li, Haobo Zhang, Bin Chen, Jiayu Zhou · 2025
Read Chain-of-Generation: Progressive Latent Diffusion for Text-Guided Molecular Design by Lingxiao Li, Haobo Zhang, Bin Chen, Jiayu Zhou (2025) 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-2511-11894.txt.
Description
Lingxiao Li and three co-authors generate molecules from written descriptions in stages. A chemist should be able to name functional groups, scaffolds, and property limits in words. Structures that honor all of them should come back. One-shot conditioning encodes the whole prompt once. It routinely drops some of the stated requirements.
Their frame, Chain-of-Generation, applies the description step by step. Each part of the prompt guides its own phase of generation. Three failures of the one-shot style are named. They include poor clarity of generated parts and missing substructures. Stepwise conditioning gives better compliance with the prompt. It also gives a clearer view of which words built which part.
Catalog record
| CREATOR | Lingxiao Li, Haobo Zhang, Bin Chen, Jiayu Zhou |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-11894