Skip to content
AI News HubLIVE
Original source2 min read

See Every Block Once: Variance-Reduced Post-Training for Masked Diffusion Language Models

Summary

Researchers propose a full-block estimator for post-training masked diffusion language models that evaluates each block exactly once, eliminating block-selection variance and achieving lower variance than random block sampling. Experiments on five benchmarks show highest accuracy on four, with an average gain of 12.79 points over the baseline.

SourcearXiv Computational LinguisticsAuthor: Yunzhen He, Hiroaki Yamagiwa, Hidetoshi Shimodaira
See Every Block Once: Variance-Reduced Post-Training for Masked Diffusion Language Models
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 13 Nov 2026]

Title:See Every Block Once: Variance-Reduced Post-Training for Masked Diffusion Language Models

View a PDF of the paper titled See Every Block Once: Variance-Reduced Post-Training for Masked Diffusion Language Models, by Yunzhen He and 2 other authors

View PDF HTML (experimental)

Abstract:Policy gradient methods are widely used to align large language model outputs with human preferences, but the log-likelihoods they require cannot be directly computed in Masked Diffusion Language Models (MDLMs). MDLM post-training therefore relies on a surrogate objective based on the Evidence Lower Bound (ELBO), and recent work estimates a blockwise variant of this objective via Monte Carlo methods to support blockwise generation. However, existing estimation schemes randomly sample the block index, so they may evaluate the same block more than once and leave some blocks unevaluated. This introduces variance arising from block selection into the estimates. In this work, we propose a full-block estimator that evaluates every block exactly once and averages the results. We theoretically show that the proposed estimator remains unbiased while eliminating the between-group variance arising from block selection, thereby achieving lower variance than the baseline for the same sample size. In experiments on five benchmarks spanning mathematics, planning, and coding, our method achieves the highest accuracy on four benchmarks and outperforms the random block sampling baseline by an average of 12.79 points.

Comments: AACL-IJCNLP 2026 Main

Subjects:

Computation and Language (cs.CL)

Cite as: arXiv:2611.17754 [cs.CL]

(or arXiv:2611.17754v1 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2611.17754

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Yunzhen He [view email] [v1] Fri, 13 Nov 2026 05:10:56 UTC (13,325 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled See Every Block Once: Variance-Reduced Post-Training for Masked Diffusion Language Models, by Yunzhen He and 2 other authors

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.CL

new | recent | 2026-11

Change to browse by:

cs

References & Citations

SciX

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

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.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • Full-block estimator evaluates each block once and averages results, remaining unbiased while removing between-group variance
  • Lower variance than random block sampling baseline for the same sample size, improving policy gradient post-training stability
  • Highest accuracy on four of five benchmarks spanning math, planning, and coding, with average 12.79-point improvement

Highlights and analysis are generated automatically and may contain errors. Check the original source.