Controlling Exploration-Exploitation in GFlowNets via Markov Chain Perspectives
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910033413931008 |
|---|---|
| author | Chen, Lin Drapeau, Samuel Shao, Fanghao Zhu, Xuekai Xue, Bo Song, Yunchong Laurière, Mathieu Lin, Zhouhan |
| author_facet | Chen, Lin Drapeau, Samuel Shao, Fanghao Zhu, Xuekai Xue, Bo Song, Yunchong Laurière, Mathieu Lin, Zhouhan |
| contents | Generative Flow Network (GFlowNet) objectives implicitly fix an equal mixing of forward and backward policies, potentially constraining the exploration-exploitation trade-off during training. By further exploring the link between GFlowNets and Markov chains, we establish an equivalence between GFlowNet objectives and Markov chain reversibility, thereby revealing the origin of such constraints, and provide a framework for adapting Markov chain properties to GFlowNets. Building on these theoretical findings, we propose $α$-GFNs, which generalize the mixing via a tunable parameter $α$. This generalization enables direct control over exploration-exploitation dynamics to enhance mode discovery capabilities, while ensuring convergence to unique flows. Across various benchmarks, including Set, Bit Sequence, and Molecule Generation, $α$-GFN objectives consistently outperform previous GFlowNet objectives, achieving up to a $10 \times$ increase in the number of discovered modes. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_01749 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Controlling Exploration-Exploitation in GFlowNets via Markov Chain Perspectives Chen, Lin Drapeau, Samuel Shao, Fanghao Zhu, Xuekai Xue, Bo Song, Yunchong Laurière, Mathieu Lin, Zhouhan Artificial Intelligence Machine Learning Generative Flow Network (GFlowNet) objectives implicitly fix an equal mixing of forward and backward policies, potentially constraining the exploration-exploitation trade-off during training. By further exploring the link between GFlowNets and Markov chains, we establish an equivalence between GFlowNet objectives and Markov chain reversibility, thereby revealing the origin of such constraints, and provide a framework for adapting Markov chain properties to GFlowNets. Building on these theoretical findings, we propose $α$-GFNs, which generalize the mixing via a tunable parameter $α$. This generalization enables direct control over exploration-exploitation dynamics to enhance mode discovery capabilities, while ensuring convergence to unique flows. Across various benchmarks, including Set, Bit Sequence, and Molecule Generation, $α$-GFN objectives consistently outperform previous GFlowNet objectives, achieving up to a $10 \times$ increase in the number of discovered modes. |
| title | Controlling Exploration-Exploitation in GFlowNets via Markov Chain Perspectives |
| topic | Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2602.01749 |