Boosted GFlowNets: Improving Exploration via Sequential Learning

Pedro Dall’Antonia, Tiago Silva, Daniel Augusto de Souza, César Lincoln Mattos, Diego Mesquita
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:2251-2259, 2026.

Abstract

Generative Flow Networks (GFlowNets) are powerful samplers for compositional objects that, by design, sample proportionally to a given non-negative reward. Nonetheless, in practice, they often struggle to explore the reward landscape evenly: trajectories toward easy-to-reach regions dominate training, while hard-to-reach modes receive vanishing or uninformative gradients, leading to poor coverage of high-reward areas. We address this imbalance with Boosted GFlowNets, a method that sequentially trains an ensemble of GFlowNets, each optimizing a residual reward that compensates for the mass already captured by previous models. This residual principle reactivates learning signals in underexplored regions and, under mild assumptions, ensures a monotone non-degradation property: adding boosters cannot worsen the learned distribution and typically improves it. Empirically, Boosted GFlowNets achieve substantially better exploration and sample diversity on multimodal synthetic benchmarks and peptide design tasks, while preserving the stability and simplicity of standard trajectory-balance training.

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-dall-antonia26a, title = { Boosted GFlowNets: Improving Exploration via Sequential Learning }, author = {Dall'Antonia, Pedro and Silva, Tiago and de Souza, Daniel Augusto and Mattos, C{\'e}sar Lincoln and Mesquita, Diego}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {2251--2259}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/dall-antonia26a/dall-antonia26a.pdf}, url = {https://proceedings.mlr.press/v300/dall-antonia26a.html}, abstract = { Generative Flow Networks (GFlowNets) are powerful samplers for compositional objects that, by design, sample proportionally to a given non-negative reward. Nonetheless, in practice, they often struggle to explore the reward landscape evenly: trajectories toward easy-to-reach regions dominate training, while hard-to-reach modes receive vanishing or uninformative gradients, leading to poor coverage of high-reward areas. We address this imbalance with Boosted GFlowNets, a method that sequentially trains an ensemble of GFlowNets, each optimizing a residual reward that compensates for the mass already captured by previous models. This residual principle reactivates learning signals in underexplored regions and, under mild assumptions, ensures a monotone non-degradation property: adding boosters cannot worsen the learned distribution and typically improves it. Empirically, Boosted GFlowNets achieve substantially better exploration and sample diversity on multimodal synthetic benchmarks and peptide design tasks, while preserving the stability and simplicity of standard trajectory-balance training. } }
Endnote
%0 Conference Paper %T Boosted GFlowNets: Improving Exploration via Sequential Learning %A Pedro Dall’Antonia %A Tiago Silva %A Daniel Augusto de Souza %A César Lincoln Mattos %A Diego Mesquita %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-dall-antonia26a %I PMLR %P 2251--2259 %U https://proceedings.mlr.press/v300/dall-antonia26a.html %V 300 %X Generative Flow Networks (GFlowNets) are powerful samplers for compositional objects that, by design, sample proportionally to a given non-negative reward. Nonetheless, in practice, they often struggle to explore the reward landscape evenly: trajectories toward easy-to-reach regions dominate training, while hard-to-reach modes receive vanishing or uninformative gradients, leading to poor coverage of high-reward areas. We address this imbalance with Boosted GFlowNets, a method that sequentially trains an ensemble of GFlowNets, each optimizing a residual reward that compensates for the mass already captured by previous models. This residual principle reactivates learning signals in underexplored regions and, under mild assumptions, ensures a monotone non-degradation property: adding boosters cannot worsen the learned distribution and typically improves it. Empirically, Boosted GFlowNets achieve substantially better exploration and sample diversity on multimodal synthetic benchmarks and peptide design tasks, while preserving the stability and simplicity of standard trajectory-balance training.
APA
Dall’Antonia, P., Silva, T., de Souza, D.A., Mattos, C.L. & Mesquita, D.. (2026). Boosted GFlowNets: Improving Exploration via Sequential Learning . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:2251-2259 Available from https://proceedings.mlr.press/v300/dall-antonia26a.html.

Related Material