@inproceedings{weng-etal-2026-q,
title = "{Q}-Align: Query-Boundary Multilingual Alignment for {LLM} Safety Guardrails",
author = "Weng, Jiaqi and
Liu, Meizhen and
Guo, Xu and
Lin, Junyu and
Li, Jinfeng and
Tao, Jialing and
Chen, YueFeng and
Xue, Hui",
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.240/",
pages = "5484--5498",
ISBN = "979-8-89176-459-0",
abstract = "Multilingual safety guardrails are essential for deployed LLMs, yet they perform poorly on low-resource languages. Prior work typically relies on large-scale multilingual safety data synthesis, which incurs substantial training costs. Cross-lingual transfer offers a cheaper alternative using limited parallel data, and recent work shows that aligning only last-token hidden states can improve multilingual safety. However, although effective for query-only (Q) risk detection, last-token alignment consistently degrades query-response (QR) detection in multilingual settings, limiting its applicability in practical guardrail systems where Q and QR serve as pre-generation and post-generation safety barriers, respectively. In this paper, we propose Q-Align, a method that decouples query intent from response evaluation by aligning only at the query boundary. Specifically, for each parallel sample, Q-Align extracts the hidden state right after the query and aligns multilingual query representations toward a confidence-weighted anchor, allowing languages with stronger intent understanding to guide transfer. We further apply language grouping based on geographic proximity to stabilize cross-lingual alignment. Experiments demonstrate that Q-Align consistently improves multilingual safety detection in both Q and QR scenarios, raising QR average F1 from 74.70{\%} to 80.79{\%} on Polyguard-Qwen, while preserving English performance and generalizing to languages unseen during training."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="weng-etal-2026-q">
<titleInfo>
<title>Q-Align: Query-Boundary Multilingual Alignment for LLM Safety Guardrails</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jiaqi</namePart>
<namePart type="family">Weng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Meizhen</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xu</namePart>
<namePart type="family">Guo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junyu</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinfeng</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jialing</namePart>
<namePart type="family">Tao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">YueFeng</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hui</namePart>
<namePart type="family">Xue</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>Multilingual safety guardrails are essential for deployed LLMs, yet they perform poorly on low-resource languages. Prior work typically relies on large-scale multilingual safety data synthesis, which incurs substantial training costs. Cross-lingual transfer offers a cheaper alternative using limited parallel data, and recent work shows that aligning only last-token hidden states can improve multilingual safety. However, although effective for query-only (Q) risk detection, last-token alignment consistently degrades query-response (QR) detection in multilingual settings, limiting its applicability in practical guardrail systems where Q and QR serve as pre-generation and post-generation safety barriers, respectively. In this paper, we propose Q-Align, a method that decouples query intent from response evaluation by aligning only at the query boundary. Specifically, for each parallel sample, Q-Align extracts the hidden state right after the query and aligns multilingual query representations toward a confidence-weighted anchor, allowing languages with stronger intent understanding to guide transfer. We further apply language grouping based on geographic proximity to stabilize cross-lingual alignment. Experiments demonstrate that Q-Align consistently improves multilingual safety detection in both Q and QR scenarios, raising QR average F1 from 74.70% to 80.79% on Polyguard-Qwen, while preserving English performance and generalizing to languages unseen during training.</abstract>
<identifier type="citekey">weng-etal-2026-q</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.240/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>5484</start>
<end>5498</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Q-Align: Query-Boundary Multilingual Alignment for LLM Safety Guardrails
%A Weng, Jiaqi
%A Liu, Meizhen
%A Guo, Xu
%A Lin, Junyu
%A Li, Jinfeng
%A Tao, Jialing
%A Chen, YueFeng
%A Xue, Hui
%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 weng-etal-2026-q
%X Multilingual safety guardrails are essential for deployed LLMs, yet they perform poorly on low-resource languages. Prior work typically relies on large-scale multilingual safety data synthesis, which incurs substantial training costs. Cross-lingual transfer offers a cheaper alternative using limited parallel data, and recent work shows that aligning only last-token hidden states can improve multilingual safety. However, although effective for query-only (Q) risk detection, last-token alignment consistently degrades query-response (QR) detection in multilingual settings, limiting its applicability in practical guardrail systems where Q and QR serve as pre-generation and post-generation safety barriers, respectively. In this paper, we propose Q-Align, a method that decouples query intent from response evaluation by aligning only at the query boundary. Specifically, for each parallel sample, Q-Align extracts the hidden state right after the query and aligns multilingual query representations toward a confidence-weighted anchor, allowing languages with stronger intent understanding to guide transfer. We further apply language grouping based on geographic proximity to stabilize cross-lingual alignment. Experiments demonstrate that Q-Align consistently improves multilingual safety detection in both Q and QR scenarios, raising QR average F1 from 74.70% to 80.79% on Polyguard-Qwen, while preserving English performance and generalizing to languages unseen during training.
%U https://aclanthology.org/2026.emnlp-main.240/
%P 5484-5498
Markdown (Informal)
[Q-Align: Query-Boundary Multilingual Alignment for LLM Safety Guardrails](https://aclanthology.org/2026.emnlp-main.240/) (Weng et al., EMNLP 2026)
ACL
- Jiaqi Weng, Meizhen Liu, Xu Guo, Junyu Lin, Jinfeng Li, Jialing Tao, YueFeng Chen, and Hui Xue. 2026. Q-Align: Query-Boundary Multilingual Alignment for LLM Safety Guardrails. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 5484–5498, Budapest, Hungary. Association for Computational Linguistics.