@inproceedings{park-etal-2026-signals,
title = "Signals from the Future: Future-aware Semi-Autoregressive Decoding for Diffusion Language Models",
author = "Park, Yangryeol and
Kang, Jeonhwi and
Lee, Kunhui and
Kim, Hyeonju and
Choi, Jinyoung and
Na, Seung-Hoon",
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.2327/",
pages = "52853--52871",
ISBN = "979-8-89176-459-0",
abstract = "Standard semi-autoregressive decoding for full-attention masked diffusion language models (MDLMs) computes predictions for future blocks but discards them until those blocks become active. When introduced as soft-future context, these predictions reveal a stability contrast: hypotheses later revised during decoding exhibit larger distributional shifts than those eventually retained. We introduce FUSA (Future-aware Semi-Autoregressive Decoding), a training-free decoder that exploits this contrast by alternating the next block between masked and soft representations and committing active-block tokens that remain stable across the two views, without additional model forward calls. On LLaDA-8B-Instruct, FUSA improves scores across reasoning and code-generation benchmarks while delivering 2.10{--} 3.94{\texttimes} wall-clock speedups over Top-1 decoding, with gains of up to 5.1 points. Additional results show that FUSA transfers to other MDLMs and remains effective under KV-cached inference and in long-form instruction following. Our code is available at https://github.com/yangryeol/FUSA."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="park-etal-2026-signals">
<titleInfo>
<title>Signals from the Future: Future-aware Semi-Autoregressive Decoding for Diffusion Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yangryeol</namePart>
<namePart type="family">Park</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeonhwi</namePart>
<namePart type="family">Kang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kunhui</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hyeonju</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinyoung</namePart>
<namePart type="family">Choi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seung-Hoon</namePart>
<namePart type="family">Na</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>Standard semi-autoregressive decoding for full-attention masked diffusion language models (MDLMs) computes predictions for future blocks but discards them until those blocks become active. When introduced as soft-future context, these predictions reveal a stability contrast: hypotheses later revised during decoding exhibit larger distributional shifts than those eventually retained. We introduce FUSA (Future-aware Semi-Autoregressive Decoding), a training-free decoder that exploits this contrast by alternating the next block between masked and soft representations and committing active-block tokens that remain stable across the two views, without additional model forward calls. On LLaDA-8B-Instruct, FUSA improves scores across reasoning and code-generation benchmarks while delivering 2.10– 3.94× wall-clock speedups over Top-1 decoding, with gains of up to 5.1 points. Additional results show that FUSA transfers to other MDLMs and remains effective under KV-cached inference and in long-form instruction following. Our code is available at https://github.com/yangryeol/FUSA.</abstract>
<identifier type="citekey">park-etal-2026-signals</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2327/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>52853</start>
<end>52871</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Signals from the Future: Future-aware Semi-Autoregressive Decoding for Diffusion Language Models
%A Park, Yangryeol
%A Kang, Jeonhwi
%A Lee, Kunhui
%A Kim, Hyeonju
%A Choi, Jinyoung
%A Na, Seung-Hoon
%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 park-etal-2026-signals
%X Standard semi-autoregressive decoding for full-attention masked diffusion language models (MDLMs) computes predictions for future blocks but discards them until those blocks become active. When introduced as soft-future context, these predictions reveal a stability contrast: hypotheses later revised during decoding exhibit larger distributional shifts than those eventually retained. We introduce FUSA (Future-aware Semi-Autoregressive Decoding), a training-free decoder that exploits this contrast by alternating the next block between masked and soft representations and committing active-block tokens that remain stable across the two views, without additional model forward calls. On LLaDA-8B-Instruct, FUSA improves scores across reasoning and code-generation benchmarks while delivering 2.10– 3.94× wall-clock speedups over Top-1 decoding, with gains of up to 5.1 points. Additional results show that FUSA transfers to other MDLMs and remains effective under KV-cached inference and in long-form instruction following. Our code is available at https://github.com/yangryeol/FUSA.
%U https://aclanthology.org/2026.emnlp-main.2327/
%P 52853-52871
Markdown (Informal)
[Signals from the Future: Future-aware Semi-Autoregressive Decoding for Diffusion Language Models](https://aclanthology.org/2026.emnlp-main.2327/) (Park et al., EMNLP 2026)
ACL