What to randomize matters: a controlled study of volume, axis and representation in low-data molecular generation
Abstract
Chemical language models can struggle to generate new molecules when trained on small datasets, where a model may reproduce training examples rather than learn a useful generative distribution. Randomized SMILES augmentation varies atom traversal, whereas Sequential Attachment-based Fragment Embedding (SAFE) also allows the order of fragment blocks to vary. How these distinct sources of representational variation affect low-data molecular generation has not been clearly studied. Here, using 1,000 seed molecules to train an autoregressive sequence model, we compare canonical SAFE with randomized SAFE at increasing corpus sizes, isolate atom-order and fragment-order axes of SAFE randomization, and compare fully randomized SAFE with randomized SMILES at molecule-by-molecule matched training corpora. Three findings emerge. First, randomization of SAFE alone raises the genuine yield of novel molecules more than 450-fold, turning a model that merely reproduces its training set into a practical molecular generator. Second, the two axes of randomization act jointly rather than additively: atom order alone reaches a tenth, and fragment order alone a seventh, of the genuine yield obtained when both vary, which is four times their combined contribution. Third, at matched corpus size and compute, randomized SAFE delivers 2.8 times the genuine yield of randomized SMILES. Welch tests with Games-Howell post-hoc comparisons resolve every difference on the primary outcomes. In the low-data regime, thus, what one randomizes rather than only how much determines whether augmentation produces new molecules or merely new spellings of the training set.
Related
- Structural Hierarchy and Geometry in Molecular Representation Learning
- Next-Generation Liposomal and Nanoparticle-Based Systems for Glioblastoma: Molecular Targeting Strategies to Overcome the Blood–Brain Barrier
- Automatic Generation of Novel Highly-Functional Molecular Sequences
- Molecular Mechanisms of the Gut-Liver-Placenta Axis in Intrahepatic Cholestasis of Pregnancy: Microbiota-Associated Pathogenesis and Next-Generation Therapeutic Strategies.
- Next-Generation Adaptogens: Molecular Mechanisms, Bioactive Compounds and Human Health Relevance of Selected Botanicals and Fungi.
BibTeX
@misc{kaushal2026what,
title = {{What to randomize matters: a controlled study of volume, axis and representation in low-data molecular generation}},
author = {Neerav Kaushal and Alisa Teige and Ajay MNV Penmatsa},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.38835},
eprint = {2610.38835},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.38835},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - What to randomize matters: a controlled study of volume, axis and representation in low-data molecular generation
AU - Neerav Kaushal
AU - Alisa Teige
AU - Ajay MNV Penmatsa
PY - 2026
DA - 2026/10/28
PB - arXiv
DO - 10.48550/arxiv.2610.38835
UR - https://arxiv.org/abs/2610.38835
AB - Chemical language models can struggle to generate new molecules when trained on small datasets, where a model may reproduce training examples rather than learn a useful generative distribution. Randomized SMILES augmentation varies atom traversal, whereas Sequential Attachment-based Fragment Embedding (SAFE) also allows the order of fragment blocks to vary. How these distinct sources of representational variation affect low-data molecular generation has not been clearly studied. Here, using 1,000 seed molecules to train an autoregressive sequence model, we compare canonical SAFE with randomized SAFE at increasing corpus sizes, isolate atom-order and fragment-order axes of SAFE randomization, and compare fully randomized SAFE with randomized SMILES at molecule-by-molecule matched training corpora. Three findings emerge. First, randomization of SAFE alone raises the genuine yield of novel molecules more than 450-fold, turning a model that merely reproduces its training set into a practical molecular generator. Second, the two axes of randomization act jointly rather than additively: atom order alone reaches a tenth, and fragment order alone a seventh, of the genuine yield obtained when both vary, which is four times their combined contribution. Third, at matched corpus size and compute, randomized SAFE delivers 2.8 times the genuine yield of randomized SMILES. Welch tests with Games-Howell post-hoc comparisons resolve every difference on the primary outcomes. In the low-data regime, thus, what one randomizes rather than only how much determines whether augmentation produces new molecules or merely new spellings of the training set.
ER -