aiqbench analysis
This paper establishes the theoretical equivalence between Generative Marginalization Models and Generative Flow Networks, introducing "Particle GFlowNets" to accelerate training in large combinatorial spaces. By extending these models to non-autoregressive processes and incorporating an automatic rejuvenation criterion based on the Gelman-Rubin statistic, the authors significantly improve convergence speeds for discrete distribution sampling. The work represents a notable advancement in the mathematical foundations of generative modeling and inference algorithms.
Particle GFlowNets: Rethinking Generative Marginalization Models
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.
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.
Machine Learning (cs.LG)
arXiv:2609.11538 [cs.LG]
(or arXiv:2609.11538v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.11538 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 224 words
Related coverage
D . A . T . E ., A LEADER IN THERMAL INNOVATION FOR CRITICAL TECHNOLOGIES , ANNOUNCES ITS PROPOSED INITIAL PUBLIC OFFERING ON EURONEXT GROWTH IN PARIS
finanznachrichten.de · Jul 30, 2026
Asymptotically Exact Negative Guidance of Diffusion Models via Positive-Unlabeled Learning
arXiv · Oct 15, 2026
Joint Bayesian Inference of Graphical Structure and Parameters with a Single Generative Flow Network
arXiv · Aug 31, 2026
CRNDiff: Count-Native Diffusion Framework via Chemical Reaction Networks
arXiv · Sep 28, 2026