@inproceedings{an-etal-2026-signflip,
title = "{S}ign{FLIP}: A Unified Model for Sign Language Translation and Generation via Stage-wise Alignment at Scale",
author = "An, Zhaoyi and
Tan, Sihan and
Hwang, Youngbae and
Nakadai, Kazuhiro and
Kawakami, Rei",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.828/",
doi = "10.18653/2026.findings-emnlp.828",
pages = "17562--17576",
ISBN = "979-8-89176-462-0",
abstract = "Sign language translation and generation share the goal of bidirectional alignment between text and sign representations. However, existing approaches either treat them as isolated tasks or are only verified on limited datasets, limiting effective modeling between modalities. In this paper, we propose SignFLIP, a unified LLM-centered framework for translation and generation. To enable bidirectional mapping between text and sign, SignFLIP adopts a symmetric architecture together with a stage-wise training strategy built on large-scale data. The shared sign{--}text representation is progressively refined: pre-alignment facilitates subsequent SLT, while the SLT-adapted representation further benefits SLG. Extensive experiments on multiple benchmarks show that SignFLIP shows competitive performance compared with task-specific models on both translation and generation tasks, as well as strong transferability to sign language recognition."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="an-etal-2026-signflip">
<titleInfo>
<title>SignFLIP: A Unified Model for Sign Language Translation and Generation via Stage-wise Alignment at Scale</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhaoyi</namePart>
<namePart type="family">An</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sihan</namePart>
<namePart type="family">Tan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Youngbae</namePart>
<namePart type="family">Hwang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kazuhiro</namePart>
<namePart type="family">Nakadai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rei</namePart>
<namePart type="family">Kawakami</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Sign language translation and generation share the goal of bidirectional alignment between text and sign representations. However, existing approaches either treat them as isolated tasks or are only verified on limited datasets, limiting effective modeling between modalities. In this paper, we propose SignFLIP, a unified LLM-centered framework for translation and generation. To enable bidirectional mapping between text and sign, SignFLIP adopts a symmetric architecture together with a stage-wise training strategy built on large-scale data. The shared sign–text representation is progressively refined: pre-alignment facilitates subsequent SLT, while the SLT-adapted representation further benefits SLG. Extensive experiments on multiple benchmarks show that SignFLIP shows competitive performance compared with task-specific models on both translation and generation tasks, as well as strong transferability to sign language recognition.</abstract>
<identifier type="citekey">an-etal-2026-signflip</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.828</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.828/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>17562</start>
<end>17576</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SignFLIP: A Unified Model for Sign Language Translation and Generation via Stage-wise Alignment at Scale
%A An, Zhaoyi
%A Tan, Sihan
%A Hwang, Youngbae
%A Nakadai, Kazuhiro
%A Kawakami, Rei
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F an-etal-2026-signflip
%X Sign language translation and generation share the goal of bidirectional alignment between text and sign representations. However, existing approaches either treat them as isolated tasks or are only verified on limited datasets, limiting effective modeling between modalities. In this paper, we propose SignFLIP, a unified LLM-centered framework for translation and generation. To enable bidirectional mapping between text and sign, SignFLIP adopts a symmetric architecture together with a stage-wise training strategy built on large-scale data. The shared sign–text representation is progressively refined: pre-alignment facilitates subsequent SLT, while the SLT-adapted representation further benefits SLG. Extensive experiments on multiple benchmarks show that SignFLIP shows competitive performance compared with task-specific models on both translation and generation tasks, as well as strong transferability to sign language recognition.
%R 10.18653/2026.findings-emnlp.828
%U https://aclanthology.org/2026.findings-emnlp.828/
%U https://doi.org/10.18653/2026.findings-emnlp.828
%P 17562-17576
Markdown (Informal)
[SignFLIP: A Unified Model for Sign Language Translation and Generation via Stage-wise Alignment at Scale](https://aclanthology.org/2026.findings-emnlp.828/) (An et al., Findings 2026)
ACL