@inproceedings{jin-etal-2026-multi,
title = "Multi-Block Diffusion Language Models",
author = "Jin, Yijie and
Xu, Jiajun and
Liu, Yuxuan and
Xu, Chenkai and
Li, Jiajun and
Yu, Kai and
Liu, Pengfei and
Deng, Zhijie",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1971/",
pages = "44855--44875",
ISBN = "979-8-89176-459-0",
abstract = "Block Diffusion Language Models (BD-LMs) support KV caching and flexible-length generation, but their native Single-Block Diffusion (SingleBD) inference leaves inter-block parallelism unused. We study Multi-Block Diffusion (MultiBD), where a running-set of consecutive blocks is refined concurrently. Existing teacher-forced BD-LMs observe only one noisy block conditioned on a clean prefix, while diffusion forcing exposes a long noisy suffix whose state distribution still differs from practical MultiBD inference. We bridge this gap with Multi-Block Diffusion Language Models (MBD-LMs), obtained by post-training BD-LMs using Multi-block Teacher Forcing (MultiTF). MultiTF trains on bounded noise-groups conditioned on clean prefixes, with heterogeneous slot-wise noise patterns that better match inference. An optimized Block Buffer further preserves prefix-cache reuse and static input shapes, translating inter-block parallelism into wall-clock acceleration. On math and code benchmarks, MBD-LLaDA2-Mini increases average Tokens Per Forward pass (TPF) from 3.47 to 6.19 while maintaining comparable accuracy (80.0{\%} versus 81.0{\%}). Combined with DMax, it reaches 9.34 TPF with a 1.02-point accuracy drop. Project Page: https://sjtu-deng-lab.github.io/mbd-lms"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jin-etal-2026-multi">
<titleInfo>
<title>Multi-Block Diffusion Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yijie</namePart>
<namePart type="family">Jin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuxuan</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chenkai</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pengfei</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhijie</namePart>
<namePart type="family">Deng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Block Diffusion Language Models (BD-LMs) support KV caching and flexible-length generation, but their native Single-Block Diffusion (SingleBD) inference leaves inter-block parallelism unused. We study Multi-Block Diffusion (MultiBD), where a running-set of consecutive blocks is refined concurrently. Existing teacher-forced BD-LMs observe only one noisy block conditioned on a clean prefix, while diffusion forcing exposes a long noisy suffix whose state distribution still differs from practical MultiBD inference. We bridge this gap with Multi-Block Diffusion Language Models (MBD-LMs), obtained by post-training BD-LMs using Multi-block Teacher Forcing (MultiTF). MultiTF trains on bounded noise-groups conditioned on clean prefixes, with heterogeneous slot-wise noise patterns that better match inference. An optimized Block Buffer further preserves prefix-cache reuse and static input shapes, translating inter-block parallelism into wall-clock acceleration. On math and code benchmarks, MBD-LLaDA2-Mini increases average Tokens Per Forward pass (TPF) from 3.47 to 6.19 while maintaining comparable accuracy (80.0% versus 81.0%). Combined with DMax, it reaches 9.34 TPF with a 1.02-point accuracy drop. Project Page: https://sjtu-deng-lab.github.io/mbd-lms</abstract>
<identifier type="citekey">jin-etal-2026-multi</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1971/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>44855</start>
<end>44875</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Multi-Block Diffusion Language Models
%A Jin, Yijie
%A Xu, Jiajun
%A Liu, Yuxuan
%A Xu, Chenkai
%A Li, Jiajun
%A Yu, Kai
%A Liu, Pengfei
%A Deng, Zhijie
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F jin-etal-2026-multi
%X Block Diffusion Language Models (BD-LMs) support KV caching and flexible-length generation, but their native Single-Block Diffusion (SingleBD) inference leaves inter-block parallelism unused. We study Multi-Block Diffusion (MultiBD), where a running-set of consecutive blocks is refined concurrently. Existing teacher-forced BD-LMs observe only one noisy block conditioned on a clean prefix, while diffusion forcing exposes a long noisy suffix whose state distribution still differs from practical MultiBD inference. We bridge this gap with Multi-Block Diffusion Language Models (MBD-LMs), obtained by post-training BD-LMs using Multi-block Teacher Forcing (MultiTF). MultiTF trains on bounded noise-groups conditioned on clean prefixes, with heterogeneous slot-wise noise patterns that better match inference. An optimized Block Buffer further preserves prefix-cache reuse and static input shapes, translating inter-block parallelism into wall-clock acceleration. On math and code benchmarks, MBD-LLaDA2-Mini increases average Tokens Per Forward pass (TPF) from 3.47 to 6.19 while maintaining comparable accuracy (80.0% versus 81.0%). Combined with DMax, it reaches 9.34 TPF with a 1.02-point accuracy drop. Project Page: https://sjtu-deng-lab.github.io/mbd-lms
%U https://aclanthology.org/2026.emnlp-main.1971/
%P 44855-44875
Markdown (Informal)
[Multi-Block Diffusion Language Models](https://aclanthology.org/2026.emnlp-main.1971/) (Jin et al., EMNLP 2026)
ACL
- Yijie Jin, Jiajun Xu, Yuxuan Liu, Chenkai Xu, Jiajun Li, Kai Yu, Pengfei Liu, and Zhijie Deng. 2026. Multi-Block Diffusion Language Models. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 44855–44875, Budapest, Hungary. Association for Computational Linguistics.