This work describes 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.
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.
It is demonstrated that the Markovian assumption can both hamper signal propagation during training and catastrophically reduce the learned sampler's expressivity due to state aliasing.
Tiago da Silva, Esmeralda S. Whitammer, S. Lahlou· 0 citations
The p(parameter)-CSMC algorithm is proposed, which incorporates both parameter learning and ancestor sampling, leading to much better mixing properties compared to (particle) Gibbs sampling in settings where strong internal correlations may challenge effective exploration.
Likelihood Score Approximation (LSA), a generative framework that keeps a pretrained unconditional model fixed and learns an observation-conditioned model that approximates the likelihood score from paired samples, is introduced.
Rostislav Makarov, Tal Peer, Danilo de Oliveira et al.· arXiv.org· 0 citations
It is shown that stochastic autoregressive learning fundamentally differs from the deterministic theory, and that CoT learning at scale $\varepsilon$ is upper-bounded by base learning at scale $\varepsilon/M^2$, whereas e2e learning at scale $\varepsilon$ is upper-bounded, up to logarithmic factors, by $(M/\varepsilon)...
Ilan Doron-Arad, Idan Mehalel, Elchanan Mossel· 1 citation
Neural Optimal Particle Filters are introduced, which learn an amortized approximation to the optimal proposal from offline simulated one-step conditioning tuples, and improve sample efficiency and distributional accuracy over standard PF baselines with modest computational overhead.
Apoorv Srivastava, Eric F. Darve· arXiv.org· 0 citations
A new paradigm that has recently emerged at the interface of machine learning and computational statistical physics: the use of generative models as tools for sampling through flexible probabilistic models that can assist the sampling of distributions known only up to a normalization constant is explored.
M. Gabrié· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 17, 2026
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.
Microsoft Research Blog· microsoft.comNov 16, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 13, 2026