@inproceedings{silchenko-2026-qebreak,
title = "{QE}break at {WMT}26: A Pre-Registered Audit of Omission and Addition Asymmetry in Machine Translation Quality Metrics",
author = "Silchenko, Maksim",
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.109/",
pages = "1745--1756",
ISBN = "979-8-89176-474-3",
abstract = "Reference-free quality estimation (QE) metrics see only the source and the hypothesis, so a translation that silently drops source content leaves nothing for them to flag. QEbreak, a pre-registered contrastive challenge set submitted to the WMT26 Automated Translation Quality Evaluation task (Subtask 4), measures this blind spot: 2,886 segments over 10 language directions, built as matched {\{}base, omission, addition{\}} triples, a fact-free verbosity ladder, and numeric-contradiction pairs as a positive control. Every hypothesis, endpoint, and equivalence margin was frozen in git-timestamped commits before any score existed. Scores from 31 systems confirm the deficit and refute our mechanism. Reference-free QE metrics under-penalize omission (family asymmetry A = {\ensuremath{-}}0.51; a current QE baseline scores the omission at or above its own base in 19.5 percent of matched contests, strictly above in 14.3), and the edit-direction by reference-availability interaction is significant ({\ensuremath{\beta}} = {\ensuremath{-}}0.069, 95{\%} CI [{\ensuremath{-}}0.081, {\ensuremath{-}}0.057]). But reference-based neural metrics are asymmetric in the same direction (A = {\ensuremath{-}}0.42), outside the pre-registered {\ensuremath{\pm}}0.30 equivalence margin, so omission under-penalization belongs to learned metrics as a class; reference access softens it but does not remove it. Every returned metric penalizes pure verbosity, and two systems score corrupted numerals above correct ones."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="silchenko-2026-qebreak">
<titleInfo>
<title>QEbreak at WMT26: A Pre-Registered Audit of Omission and Addition Asymmetry in Machine Translation Quality Metrics</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maksim</namePart>
<namePart type="family">Silchenko</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>Reference-free quality estimation (QE) metrics see only the source and the hypothesis, so a translation that silently drops source content leaves nothing for them to flag. QEbreak, a pre-registered contrastive challenge set submitted to the WMT26 Automated Translation Quality Evaluation task (Subtask 4), measures this blind spot: 2,886 segments over 10 language directions, built as matched {base, omission, addition} triples, a fact-free verbosity ladder, and numeric-contradiction pairs as a positive control. Every hypothesis, endpoint, and equivalence margin was frozen in git-timestamped commits before any score existed. Scores from 31 systems confirm the deficit and refute our mechanism. Reference-free QE metrics under-penalize omission (family asymmetry A = \ensuremath-0.51; a current QE baseline scores the omission at or above its own base in 19.5 percent of matched contests, strictly above in 14.3), and the edit-direction by reference-availability interaction is significant (\ensuremathβ = \ensuremath-0.069, 95% CI [\ensuremath-0.081, \ensuremath-0.057]). But reference-based neural metrics are asymmetric in the same direction (A = \ensuremath-0.42), outside the pre-registered \ensuremath\pm0.30 equivalence margin, so omission under-penalization belongs to learned metrics as a class; reference access softens it but does not remove it. Every returned metric penalizes pure verbosity, and two systems score corrupted numerals above correct ones.</abstract>
<identifier type="citekey">silchenko-2026-qebreak</identifier>
<location>
<url>https://aclanthology.org/2026.wmt-main.109/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>1745</start>
<end>1756</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T QEbreak at WMT26: A Pre-Registered Audit of Omission and Addition Asymmetry in Machine Translation Quality Metrics
%A Silchenko, Maksim
%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 silchenko-2026-qebreak
%X Reference-free quality estimation (QE) metrics see only the source and the hypothesis, so a translation that silently drops source content leaves nothing for them to flag. QEbreak, a pre-registered contrastive challenge set submitted to the WMT26 Automated Translation Quality Evaluation task (Subtask 4), measures this blind spot: 2,886 segments over 10 language directions, built as matched {base, omission, addition} triples, a fact-free verbosity ladder, and numeric-contradiction pairs as a positive control. Every hypothesis, endpoint, and equivalence margin was frozen in git-timestamped commits before any score existed. Scores from 31 systems confirm the deficit and refute our mechanism. Reference-free QE metrics under-penalize omission (family asymmetry A = \ensuremath-0.51; a current QE baseline scores the omission at or above its own base in 19.5 percent of matched contests, strictly above in 14.3), and the edit-direction by reference-availability interaction is significant (\ensuremathβ = \ensuremath-0.069, 95% CI [\ensuremath-0.081, \ensuremath-0.057]). But reference-based neural metrics are asymmetric in the same direction (A = \ensuremath-0.42), outside the pre-registered \ensuremath\pm0.30 equivalence margin, so omission under-penalization belongs to learned metrics as a class; reference access softens it but does not remove it. Every returned metric penalizes pure verbosity, and two systems score corrupted numerals above correct ones.
%U https://aclanthology.org/2026.wmt-main.109/
%P 1745-1756Markdown (Informal)
[QEbreak at WMT26: A Pre-Registered Audit of Omission and Addition Asymmetry in Machine Translation Quality Metrics](https://aclanthology.org/2026.wmt-main.109/) (Silchenko, WMT 2026)
ACL