TUBE: Tangent Upper Bound on Evidence for Discrete Diffusion Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ivanov, Arseny, Kholkin, Sergei, Gromadskii, Vladislav, Ksenofontov, Grigoriy, Oseledets, Ivan, Korotin, Alexander
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917526868328448
author Ivanov, Arseny
Kholkin, Sergei
Gromadskii, Vladislav
Ksenofontov, Grigoriy
Oseledets, Ivan
Korotin, Alexander
author_facet Ivanov, Arseny
Kholkin, Sergei
Gromadskii, Vladislav
Ksenofontov, Grigoriy
Oseledets, Ivan
Korotin, Alexander
contents Log-likelihood is a standard metric for evaluating generative models. Unfortunately, in contrast to autoregressive models (ARMs), discrete diffusion models generally do not admit exact computation of this quantity. Existing evaluations, therefore, rely on the evidence lower bound (ELBO), leaving unclear how much higher the true value may be. We address this by introducing the Tangent Upper Bound on Evidence (TUBE), a variational upper bound on log-likelihood that admits an unbiased Monte Carlo estimator. Our TUBE extends across latent-variable models, including masked diffusion models (MDMs), any-order ARMs (AO-ARMs), and block variants of both. Applied to block MDMs and block AO-ARMs, TUBE reveals our key empirical finding that these models lie strictly below the exact ARM baseline, showing that ARMs still dominate in likelihood.
format Preprint
id arxiv_https___arxiv_org_abs_2605_24292
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle TUBE: Tangent Upper Bound on Evidence for Discrete Diffusion Language Models
Ivanov, Arseny
Kholkin, Sergei
Gromadskii, Vladislav
Ksenofontov, Grigoriy
Oseledets, Ivan
Korotin, Alexander
Machine Learning
Log-likelihood is a standard metric for evaluating generative models. Unfortunately, in contrast to autoregressive models (ARMs), discrete diffusion models generally do not admit exact computation of this quantity. Existing evaluations, therefore, rely on the evidence lower bound (ELBO), leaving unclear how much higher the true value may be. We address this by introducing the Tangent Upper Bound on Evidence (TUBE), a variational upper bound on log-likelihood that admits an unbiased Monte Carlo estimator. Our TUBE extends across latent-variable models, including masked diffusion models (MDMs), any-order ARMs (AO-ARMs), and block variants of both. Applied to block MDMs and block AO-ARMs, TUBE reveals our key empirical finding that these models lie strictly below the exact ARM baseline, showing that ARMs still dominate in likelihood.
title TUBE: Tangent Upper Bound on Evidence for Discrete Diffusion Language Models
topic Machine Learning
url https://arxiv.org/abs/2605.24292