Why do Large Language Models Fail in Low-resource Translation? Unraveling the Token Dynamics of Large Language Models for Machine Translation

Shenbin Qian, Yves Scherrer


Abstract
Large Language Models (LLMs) have recently demonstrated strong performance in machine translation (MT). However, most prior work focuses on improving or benchmarking translation quality, offering limited insight into when and why LLM-based translation fails. In this work, we systematically analyze failure modes of LLMs in MT by evaluating 15 models, including four reasoning LLMs, across 22 language pairs (LPs) with varying resource levels. We find that non-English-centric LPs consistently yield lower COMET scores than English-centric pairs. To investigate the underlying causes, we introduce Token Activation Rate (TAR), a metric that captures how effectively a model utilizes language-specific tokens in its vocabulary during generation. We validate TAR as a proxy for language representation using models with known language distributions in the training data, and show that lower TAR is strongly associated with poorer translation performance. Furthermore, reasoning LLMs tend to generate more tokens when translating into low-TAR languages, suggesting a compensatory mechanism, although its impact on translation quality varies across models. Overall, our findings emphasize the importance of token-level dynamics in understanding MT performance of LLMs.
Anthology ID:
2026.eamt-1.4
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:
22–47
Language:
URL:
https://aclanthology.org/2026.eamt-1.4/
DOI:
Bibkey:
Cite (ACL):
Shenbin Qian and Yves Scherrer. 2026. Why do Large Language Models Fail in Low-resource Translation? Unraveling the Token Dynamics of Large Language Models for Machine Translation. In Proceedings of the 26th Annual Conference of the European Association for Machine Translation (Volume 1), pages 22–47, Tilburg, The Netherlands. European Association for Machine Translation.
Cite (Informal):
Why do Large Language Models Fail in Low-resource Translation? Unraveling the Token Dynamics of Large Language Models for Machine Translation (Qian & Scherrer, EAMT 2026)
Copy Citation:
PDF:
https://aclanthology.org/2026.eamt-1.4.pdf