@inproceedings{nguyen-tri-etal-2025-diffusion,
title = "Diffusion Directed Acyclic Transformer for Non-Autoregressive Machine Translation",
author = "Nguyen-Tri, Quan and
Tran, Cong Dao and
Thanh-Tung, Hoang",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.acl-short.64/",
doi = "10.18653/v1/2025.acl-short.64",
pages = "814--828",
ISBN = "979-8-89176-252-7",
abstract = "Non-autoregressive transformers (NATs) predict entire sequences in parallel to reduce decoding latency, but they often encounter performance challenges due to the multi-modality problem. A recent advancement, the Directed Acyclic Transformer (DAT), addresses this issue by capturing multiple translation modalities to paths in a Directed Acyclic Graph (DAG). However, the collaboration with the latent variable introduced through the Glancing training (GLAT) is crucial for DAT to attain state-of-the-art performance. In this paper, we introduce Diffusion Directed Acyclic Transformer (Diff-DAT), which serves as an alternative to GLAT as a latent variable introduction for DAT. Diff-DAT offers two significant benefits over the previous approach. Firstly, it establishes a stronger alignment between training and inference. Secondly, it facilitates a more flexible tradeoff between quality and latency."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nguyen-tri-etal-2025-diffusion">
<titleInfo>
<title>Diffusion Directed Acyclic Transformer for Non-Autoregressive Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Quan</namePart>
<namePart type="family">Nguyen-Tri</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Cong</namePart>
<namePart type="given">Dao</namePart>
<namePart type="family">Tran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hoang</namePart>
<namePart type="family">Thanh-Tung</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-252-7</identifier>
</relatedItem>
<abstract>Non-autoregressive transformers (NATs) predict entire sequences in parallel to reduce decoding latency, but they often encounter performance challenges due to the multi-modality problem. A recent advancement, the Directed Acyclic Transformer (DAT), addresses this issue by capturing multiple translation modalities to paths in a Directed Acyclic Graph (DAG). However, the collaboration with the latent variable introduced through the Glancing training (GLAT) is crucial for DAT to attain state-of-the-art performance. In this paper, we introduce Diffusion Directed Acyclic Transformer (Diff-DAT), which serves as an alternative to GLAT as a latent variable introduction for DAT. Diff-DAT offers two significant benefits over the previous approach. Firstly, it establishes a stronger alignment between training and inference. Secondly, it facilitates a more flexible tradeoff between quality and latency.</abstract>
<identifier type="citekey">nguyen-tri-etal-2025-diffusion</identifier>
<identifier type="doi">10.18653/v1/2025.acl-short.64</identifier>
<location>
<url>https://aclanthology.org/2025.acl-short.64/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>814</start>
<end>828</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Diffusion Directed Acyclic Transformer for Non-Autoregressive Machine Translation
%A Nguyen-Tri, Quan
%A Tran, Cong Dao
%A Thanh-Tung, Hoang
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-252-7
%F nguyen-tri-etal-2025-diffusion
%X Non-autoregressive transformers (NATs) predict entire sequences in parallel to reduce decoding latency, but they often encounter performance challenges due to the multi-modality problem. A recent advancement, the Directed Acyclic Transformer (DAT), addresses this issue by capturing multiple translation modalities to paths in a Directed Acyclic Graph (DAG). However, the collaboration with the latent variable introduced through the Glancing training (GLAT) is crucial for DAT to attain state-of-the-art performance. In this paper, we introduce Diffusion Directed Acyclic Transformer (Diff-DAT), which serves as an alternative to GLAT as a latent variable introduction for DAT. Diff-DAT offers two significant benefits over the previous approach. Firstly, it establishes a stronger alignment between training and inference. Secondly, it facilitates a more flexible tradeoff between quality and latency.
%R 10.18653/v1/2025.acl-short.64
%U https://aclanthology.org/2025.acl-short.64/
%U https://doi.org/10.18653/v1/2025.acl-short.64
%P 814-828
Markdown (Informal)
[Diffusion Directed Acyclic Transformer for Non-Autoregressive Machine Translation](https://aclanthology.org/2025.acl-short.64/) (Nguyen-Tri et al., ACL 2025)
ACL