A* Decoding for Machine Translation in LLMs - SRPOL Participation in WMT2025

Adam Dobrowolski, Paweł Przewłocki, Paweł Przybysz, Marcin Szymański, Dawid Siwicki


Abstract
SRPOL team submission to WMT2025 introduces innovative approach using A* (A-star) algorithm of decoding in EuroLLM which gives diverse set of translation hypotheses. Subsequent reranking by Comet-QE and NLLB chooses the best of the diversed hypotheses which gives significant improvement of translation quality. The A* algorithm can be applied to decoding in any LLMs or other translation models. The experiment shows that by using free, openly accessible MT models you can achieve translation quality of the best online translators and LLMs using just a PC under your desk.
Anthology ID:
2025.wmt-1.42
Volume:
Proceedings of the Tenth Conference on Machine Translation
Month:
November
Year:
2025
Address:
Suzhou, China
Editors:
Barry Haddow, Tom Kocmi, Philipp Koehn, Christof Monz
Venue:
WMT
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
666–670
Language:
URL:
https://preview.aclanthology.org/author-page-yu-wang-polytechnic/2025.wmt-1.42/
DOI:
10.18653/v1/2025.wmt-1.42
Bibkey:
Cite (ACL):
Adam Dobrowolski, Paweł Przewłocki, Paweł Przybysz, Marcin Szymański, and Dawid Siwicki. 2025. A* Decoding for Machine Translation in LLMs - SRPOL Participation in WMT2025. In Proceedings of the Tenth Conference on Machine Translation, pages 666–670, Suzhou, China. Association for Computational Linguistics.
Cite (Informal):
A* Decoding for Machine Translation in LLMs - SRPOL Participation in WMT2025 (Dobrowolski et al., WMT 2025)
Copy Citation:
PDF:
https://preview.aclanthology.org/author-page-yu-wang-polytechnic/2025.wmt-1.42.pdf