@inproceedings{krause-etal-2026-probe,
title = "Probe Before you Answer! Answerability Probing for Abstention in Cross-Lingual Context-Based Question Answering",
author = "Krause, Lea and
Schouten, Stefan and
Van Der Meer, Michiel and
Vossen, Piek",
editor = "Haddow, Barry and
Kocmi, Tom and
Koehn, Philipp and
Monz, Christof",
booktitle = "Proceedings of the Eleventh Conference on Machine Translation",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.wmt-main.129/",
pages = "1933--1948",
ISBN = "979-8-89176-474-3",
abstract = "We separate answering from abstention in a context-based question answering task, where a language model answers a question about a document in a possibly different language or abstains when the document contains no answer. A LoRA adapter on Qwen3.5-9B, trained only on answerable questions, generates the answer, while a linear probe on the frozen base model{'}s activations decides answerability independently and returns the required refusal string when it predicts none exists. This isolates the model{'}s answerability signal from its instruction-following. The probe{'}s ranking of answerable against unanswerable questions transfers across languages, corpora, and negative constructions it was not trained on, but its decision threshold does not; training on in-domain multilingual data, rather than recalibrating the threshold, closes most of that gap. For the adapter, how the answer target is constructed trades off automatic correctness metrics against well-formedness and answer length."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="krause-etal-2026-probe">
<titleInfo>
<title>Probe Before you Answer! Answerability Probing for Abstention in Cross-Lingual Context-Based Question Answering</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lea</namePart>
<namePart type="family">Krause</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Stefan</namePart>
<namePart type="family">Schouten</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michiel</namePart>
<namePart type="family">Van Der Meer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Piek</namePart>
<namePart type="family">Vossen</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 Eleventh Conference on Machine Translation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Barry</namePart>
<namePart type="family">Haddow</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tom</namePart>
<namePart type="family">Kocmi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Philipp</namePart>
<namePart type="family">Koehn</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Christof</namePart>
<namePart type="family">Monz</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">979-8-89176-474-3</identifier>
</relatedItem>
<abstract>We separate answering from abstention in a context-based question answering task, where a language model answers a question about a document in a possibly different language or abstains when the document contains no answer. A LoRA adapter on Qwen3.5-9B, trained only on answerable questions, generates the answer, while a linear probe on the frozen base model’s activations decides answerability independently and returns the required refusal string when it predicts none exists. This isolates the model’s answerability signal from its instruction-following. The probe’s ranking of answerable against unanswerable questions transfers across languages, corpora, and negative constructions it was not trained on, but its decision threshold does not; training on in-domain multilingual data, rather than recalibrating the threshold, closes most of that gap. For the adapter, how the answer target is constructed trades off automatic correctness metrics against well-formedness and answer length.</abstract>
<identifier type="citekey">krause-etal-2026-probe</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.129/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>1933</start>
<end>1948</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Probe Before you Answer! Answerability Probing for Abstention in Cross-Lingual Context-Based Question Answering
%A Krause, Lea
%A Schouten, Stefan
%A Van Der Meer, Michiel
%A Vossen, Piek
%Y Haddow, Barry
%Y Kocmi, Tom
%Y Koehn, Philipp
%Y Monz, Christof
%S Proceedings of the Eleventh Conference on Machine Translation
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-474-3
%F krause-etal-2026-probe
%X We separate answering from abstention in a context-based question answering task, where a language model answers a question about a document in a possibly different language or abstains when the document contains no answer. A LoRA adapter on Qwen3.5-9B, trained only on answerable questions, generates the answer, while a linear probe on the frozen base model’s activations decides answerability independently and returns the required refusal string when it predicts none exists. This isolates the model’s answerability signal from its instruction-following. The probe’s ranking of answerable against unanswerable questions transfers across languages, corpora, and negative constructions it was not trained on, but its decision threshold does not; training on in-domain multilingual data, rather than recalibrating the threshold, closes most of that gap. For the adapter, how the answer target is constructed trades off automatic correctness metrics against well-formedness and answer length.
%U https://aclanthology.org/2026.wmt-main.129/
%P 1933-1948
Markdown (Informal)
[Probe Before you Answer! Answerability Probing for Abstention in Cross-Lingual Context-Based Question Answering](https://aclanthology.org/2026.wmt-main.129/) (Krause et al., WMT 2026)
ACL