Synthesizable Molecular Generation via Soft-constrained GFlowNets with Rich Chemical Priors

Hyeonah Kim, Minsu Kim, Celine Roget, Dionessa Biton, Louis Vaillancourt, Yves V. Brun, Yoshua Bengio, Alex Hernández-Garcı́a
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:57977-58001, 2026.

Abstract

The application of generative models for experimental drug discovery campaigns is severely limited by the difficulty of designing molecules de novo that can be synthesized in practice. Previous works have leveraged Generative Flow Networks (GFlowNets) to impose hard synthesizability constraints through the design of state and action spaces based on predefined reaction templates and building blocks. Despite the promising prospects of this approach, it currently lacks flexibility and scalability. As an alternative, we propose S3-GFN, which generates synthesizable SMILES molecules via simple soft regularization of a sequence-based GFlowNet. Our approach leverages rich molecular priors learned from large-scale SMILES corpora to steer molecular generation towards high-reward, synthesizable chemical spaces. The model induces constraints through off-policy replay training with a contrastive learning signal based on separate buffers of synthesizable and unsynthesizable samples. Our experiments show that S3-GFN learns to generate synthesizable molecules ($\geq 95$%) with higher rewards in diverse tasks.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-kim26ad, title = {Synthesizable Molecular Generation via Soft-constrained {GF}low{N}ets with Rich Chemical Priors}, author = {Kim, Hyeonah and Kim, Minsu and Roget, Celine and Biton, Dionessa and Vaillancourt, Louis and Brun, Yves V. and Bengio, Yoshua and Hern\'{a}ndez-Garc\'{\i}a, Alex}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {57977--58001}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/kim26ad/kim26ad.pdf}, url = {https://proceedings.mlr.press/v306/kim26ad.html}, abstract = {The application of generative models for experimental drug discovery campaigns is severely limited by the difficulty of designing molecules de novo that can be synthesized in practice. Previous works have leveraged Generative Flow Networks (GFlowNets) to impose hard synthesizability constraints through the design of state and action spaces based on predefined reaction templates and building blocks. Despite the promising prospects of this approach, it currently lacks flexibility and scalability. As an alternative, we propose S3-GFN, which generates synthesizable SMILES molecules via simple soft regularization of a sequence-based GFlowNet. Our approach leverages rich molecular priors learned from large-scale SMILES corpora to steer molecular generation towards high-reward, synthesizable chemical spaces. The model induces constraints through off-policy replay training with a contrastive learning signal based on separate buffers of synthesizable and unsynthesizable samples. Our experiments show that S3-GFN learns to generate synthesizable molecules ($\geq 95$%) with higher rewards in diverse tasks.} }
Endnote
%0 Conference Paper %T Synthesizable Molecular Generation via Soft-constrained GFlowNets with Rich Chemical Priors %A Hyeonah Kim %A Minsu Kim %A Celine Roget %A Dionessa Biton %A Louis Vaillancourt %A Yves V. Brun %A Yoshua Bengio %A Alex Hernández-Garcı́a %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-kim26ad %I PMLR %P 57977--58001 %U https://proceedings.mlr.press/v306/kim26ad.html %V 306 %X The application of generative models for experimental drug discovery campaigns is severely limited by the difficulty of designing molecules de novo that can be synthesized in practice. Previous works have leveraged Generative Flow Networks (GFlowNets) to impose hard synthesizability constraints through the design of state and action spaces based on predefined reaction templates and building blocks. Despite the promising prospects of this approach, it currently lacks flexibility and scalability. As an alternative, we propose S3-GFN, which generates synthesizable SMILES molecules via simple soft regularization of a sequence-based GFlowNet. Our approach leverages rich molecular priors learned from large-scale SMILES corpora to steer molecular generation towards high-reward, synthesizable chemical spaces. The model induces constraints through off-policy replay training with a contrastive learning signal based on separate buffers of synthesizable and unsynthesizable samples. Our experiments show that S3-GFN learns to generate synthesizable molecules ($\geq 95$%) with higher rewards in diverse tasks.
APA
Kim, H., Kim, M., Roget, C., Biton, D., Vaillancourt, L., Brun, Y.V., Bengio, Y. & Hernández-Garcı́a, A.. (2026). Synthesizable Molecular Generation via Soft-constrained GFlowNets with Rich Chemical Priors. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:57977-58001 Available from https://proceedings.mlr.press/v306/kim26ad.html.

Related Material