<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="popov-etal-2026-beyond">
    <titleInfo>
        <title>Beyond Prompts: A Systematic Study of LLM-Based Machine Translation Evaluators</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Dmitry</namePart>
        <namePart type="family">Popov</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Roman</namePart>
        <namePart type="family">Bokhyan</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ekaterina</namePart>
        <namePart type="family">Enikeeva</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Artem</namePart>
        <namePart type="family">Mekhraliev</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Stepan</namePart>
        <namePart type="family">Vysotsky</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Nikolay</namePart>
        <namePart type="family">Karpachev</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>LLM-based evaluators have rapidly become a standard tool for machine translation evaluation, yet the design choices that shape their behavior remain poorly understood. We present a systematic study of how instruction design, few-shot demonstrations, backbone model choice, and multi-step workflows affect span-level and ranking-based MT evaluation. We collect new expert RATE annotations on WMT24 and WMT25 English–Russian translations, with transfer validation on MQM- and ESA-annotated benchmarks across five additional language pairs. Our experiments reveal four robust findings: instruction structure shapes evaluator behavior, high-quality demonstrations matter more than quantity, frontier backbones exhibit distinct evaluator profiles (recall-, precision-, or ranking-oriented), and fusing outputs across diverse backbones outperforms single-evaluator multi-call setups. Although evaluator design exposes a large combinatorial space, several discovered design principles transfer consistently across annotation protocols and language pairs. We synthesize these findings into ORBIT, a family of LLM-based MT evaluators. Our strongest configurations achieve expert-level performance in error span detection and in Accuracy and Fluency ranking at roughly 20× lower cost than expert human annotation.</abstract>
    <identifier type="citekey">popov-etal-2026-beyond</identifier>
    <identifier type="doi">10.18653/2026.findings-emnlp.2404</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.2404/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>51648</start>
            <end>51664</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Beyond Prompts: A Systematic Study of LLM-Based Machine Translation Evaluators
%A Popov, Dmitry
%A Bokhyan, Roman
%A Enikeeva, Ekaterina
%A Mekhraliev, Artem
%A Vysotsky, Stepan
%A Karpachev, Nikolay
%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 popov-etal-2026-beyond
%X LLM-based evaluators have rapidly become a standard tool for machine translation evaluation, yet the design choices that shape their behavior remain poorly understood. We present a systematic study of how instruction design, few-shot demonstrations, backbone model choice, and multi-step workflows affect span-level and ranking-based MT evaluation. We collect new expert RATE annotations on WMT24 and WMT25 English–Russian translations, with transfer validation on MQM- and ESA-annotated benchmarks across five additional language pairs. Our experiments reveal four robust findings: instruction structure shapes evaluator behavior, high-quality demonstrations matter more than quantity, frontier backbones exhibit distinct evaluator profiles (recall-, precision-, or ranking-oriented), and fusing outputs across diverse backbones outperforms single-evaluator multi-call setups. Although evaluator design exposes a large combinatorial space, several discovered design principles transfer consistently across annotation protocols and language pairs. We synthesize these findings into ORBIT, a family of LLM-based MT evaluators. Our strongest configurations achieve expert-level performance in error span detection and in Accuracy and Fluency ranking at roughly 20× lower cost than expert human annotation.
%R 10.18653/2026.findings-emnlp.2404
%U https://aclanthology.org/2026.findings-emnlp.2404/
%U https://doi.org/10.18653/2026.findings-emnlp.2404
%P 51648-51664

Markdown (Informal)

[Beyond Prompts: A Systematic Study of LLM-Based Machine Translation Evaluators](https://aclanthology.org/2026.findings-emnlp.2404/) (Popov et al., Findings 2026)

• Beyond Prompts: A Systematic Study of LLM-Based Machine Translation Evaluators (Popov et al., Findings 2026)

ACL

• Dmitry Popov, Roman Bokhyan, Ekaterina Enikeeva, Artem Mekhraliev, Stepan Vysotsky, and Nikolay Karpachev. 2026. Beyond Prompts: A Systematic Study of LLM-Based Machine Translation Evaluators. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 51648–51664, Budapest, Hungary. Association for Computational Linguistics.