LLM-as-a-Jury for Machine Translation Publishability Assessment

Alex Yanishevsky, Olivia Norris


Abstract
We propose an LLM-as-a-Jury framework for determining machine translation publishability, aggregating judgments from multiple large language models via logistic regression rather than relying on a single judge. Publishability is defined as the absence of major or critical errors—those that render a translation unsuitable for public release without human post-editing. We compare three evaluation frameworks: a generic Edit Effort Estimation (EEE) prompt based on lexical accuracy, grammatical correctness and semantic coherence, a generic Linguistic Quality Assurance (LQA) prompt based on the MQM error taxonomy, and a purpose-built Publishability prompt optimized via DSPy and augmented with domain-specific fine-tuning. Experiments across three domains and nine language pairs show that (i) the jury ensemble matches or outperforms the best individual juror in nearly every condition, (ii) EEE and LQA juries are competitive with and occasionally exceed the Publishability jury on macro-F1, (iii) the Publishability framework offers stronger precision and a more favorable error correction asymmetry, and (iv) domain-specific fine-tuning yields substantial recall gains in client-heavy domains. These results support the viability of fully automated publishability determination in enterprise MT workflows.
Anthology ID:
2026.eamt-1.7
Volume:
Proceedings of the 26th Annual Conference of the European Association for Machine Translation (Volume 1)
Month:
June
Year:
2026
Address:
Tilburg, The Netherlands
Editors:
Dimitar Shterionov, Eva Vanmassenhove, Mirella De Sisto, Fred Blain, Javad Pourmostafa Roshan Sharami, Lisa Lepp, Chiara Manna, Argentina Anna Rescigno, Alina Karakanta, Ayla Rigouts Terryn, Manuel Lardelli, Natalia Resende, Elena Murgolo, Janiça Hackenbuchner, Anna Zaretskaya, Miquel Esplà-Gomis, Thierry Etchegoyhen, Dagmar Gromann, Rachel Bawden, Barry Haddow, Sara Szoc, Mikel Forcada, Helena Moniz
Venue:
EAMT
SIG:
Publisher:
European Association for Machine Translation
Note:
Pages:
76–84
Language:
URL:
https://aclanthology.org/2026.eamt-1.7/
DOI:
Bibkey:
Cite (ACL):
Alex Yanishevsky and Olivia Norris. 2026. LLM-as-a-Jury for Machine Translation Publishability Assessment. In Proceedings of the 26th Annual Conference of the European Association for Machine Translation (Volume 1), pages 76–84, Tilburg, The Netherlands. European Association for Machine Translation.
Cite (Informal):
LLM-as-a-Jury for Machine Translation Publishability Assessment (Yanishevsky & Norris, EAMT 2026)
Copy Citation:
PDF:
https://aclanthology.org/2026.eamt-1.7.pdf