Pushing Biomolecular Utility-Diversity Frontiers with Supergroup Relative Policy Optimization
Abstract
Biomolecular generators are often adapted with reward feedback to improve task-specific utility, but pushing utility alone can concentrate generation on a narrow family of candidates. Maintaining diversity is difficult because sample diversity is a set-level property. We introduce Supergroup Relative Policy Optimization (SGRPO), a flexible GRPO-style framework that directly constructs rewards from set-level diversity. For each condition, SGRPO samples a supergroup of candidate sets, compares their diversity under the same condition, and redistributes the group diversity reward to individual rollouts through leave-one-out diversity contributions before combining it with rollout-level utility. This design decouples SGRPO from a particular generator, utility reward, or diversity metric, and allows instantiation with different GRPO-style approaches. We evaluate SGRPO on three main settings: de novo small-molecule design, pocket-based small-molecule design, and de novo protein design, instantiating it with both GRPO and Coupled-GRPO across autoregressive and discrete diffusion generators. Across decoding sweeps, SGRPO expands the utility-diversity Pareto frontier and achieves the best frontier-level metrics relative to baselines. Our analyses further show that direct set-level diversity rewards remain effective with small groups and help preserve broader generation-distribution coverage during post-training. The code is available at https://github.com/IDEA-XL/SGRPO.
Related
- Predictive all-atom simulations of disordered proteins and biomolecular condensates through osmometry-guided force-field optimization
- Mapping the frontiers of genomic disease surveillance research in the ASEAN region: scientometric insights for precision public health and next-generation epidemic intelligence
- Molecular Grammar of Intrinsically Disordered Regions: How Sequence and Context Shape Biomolecular Condensation and Disease
- GenAI-Net: A generative AI framework for automated biomolecular network design
- Long-range transport of biomolecular condensates by Marangoni effect
BibTeX
@misc{ye2026pushing,
title = {{Pushing Biomolecular Utility-Diversity Frontiers with Supergroup Relative Policy Optimization}},
author = {Xinwu Ye and He Cao and Hao Li and Bin Feng and Zijing Liu and Xiangru Tang and Yu Li and Shenghua Gao},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2605.08659},
eprint = {2605.08659},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.08659},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Pushing Biomolecular Utility-Diversity Frontiers with Supergroup Relative Policy Optimization
AU - Xinwu Ye
AU - He Cao
AU - Hao Li
AU - Bin Feng
AU - Zijing Liu
AU - Xiangru Tang
AU - Yu Li
AU - Shenghua Gao
PY - 2026
DA - 2026/10/21
PB - arXiv
DO - 10.48550/arxiv.2605.08659
UR - https://arxiv.org/abs/2605.08659
AB - Biomolecular generators are often adapted with reward feedback to improve task-specific utility, but pushing utility alone can concentrate generation on a narrow family of candidates. Maintaining diversity is difficult because sample diversity is a set-level property. We introduce Supergroup Relative Policy Optimization (SGRPO), a flexible GRPO-style framework that directly constructs rewards from set-level diversity. For each condition, SGRPO samples a supergroup of candidate sets, compares their diversity under the same condition, and redistributes the group diversity reward to individual rollouts through leave-one-out diversity contributions before combining it with rollout-level utility. This design decouples SGRPO from a particular generator, utility reward, or diversity metric, and allows instantiation with different GRPO-style approaches. We evaluate SGRPO on three main settings: de novo small-molecule design, pocket-based small-molecule design, and de novo protein design, instantiating it with both GRPO and Coupled-GRPO across autoregressive and discrete diffusion generators. Across decoding sweeps, SGRPO expands the utility-diversity Pareto frontier and achieves the best frontier-level metrics relative to baselines. Our analyses further show that direct set-level diversity rewards remain effective with small groups and help preserve broader generation-distribution coverage during post-training. The code is available at https://github.com/IDEA-XL/SGRPO.
ER -