Particle GFlowNets: Rethinking Generative Marginalization Models

A new arXiv paper (2609.11538v1) shows that Generative Marginalization Models (MaMs) are equivalent to Generative Flow Networks (GFlowNets), and introduces Particle GFlowNets, which extends MaMs' sampling strategy to non-autoregressive generative processes. The method adds an automatic criterion for full-state rejuvenation of the Gibbs sampler derived from the Gelman-Rubin statistic, which the authors report markedly accelerates training convergence in large combinatorial spaces.

arXiv:2609.11538v1 Announce Type: new Abstract: Generative Marginalization Models MaMs have been recently introduced as efficient neural sampling models for any-order autoregressive modelling of discrete distributions. By learning both the marginal and conditional probabilities of a persistent-block Gibbs sampler, MaMs enable fast posterior evaluation with a single neural network forward pass. While prior work has considered MaMs to be distinct from Generative Flow Networks GFlowNets , a well-established paradigm for inference in discrete stochastic models, we show that they are equivalent. Then, we also extend MaMs' sampling strategy to non-autoregressive generative processes. In particular, we describe an automatic criterion for full-state rejuvenation of the Gibbs sampler, derived from the Gelman-Rubin statistic, which plays a key role in speeding up learning convergence. Our experiments show that our method, called Particle GFlowNets, markedly accelerates training in large combinatorial spaces.