TUMHN@WMT2026: LLMs With Limited Resources for Slavic Languages
MCML Authors
Tsedeniya Kinfe Temesgen
Sinan Kurtyigit
Alexander Fraser
Prof. Dr.
Core PI
Abstract
Tsedeniya Kinfe Temesgen
Sinan Kurtyigit
Alexander Fraser
Prof. Dr.
Core PI
Abstract
We describe the TUMHN team’s systems for the WMT26 Shared Tasks on LLMs with Limited Resources for Slavic Languages. We participated in the Ukrainian track, covering machine translation, question answering, spell checking, grammar checking, and math reasoning tasks. We restructured our training dataset into a chatstyle format and fine-tuned the Qwen3.5-2B model for the Ukrainian track in a multitask learning setting. Our model outperforms the baseline on three of the five tasks, with significant improvements on Question Answering, Spell Checking, and Grammar Checking. In Machine Translation, the model shows a slight improvement for the English–Ukrainian direction. For Spell Checking and Grammar Checking specifically, the largest gains were observed in spell or grammatical error detection as opposed to correction. In Math Reasoning, however, our model does not outperform the baseline. The fine-tuned model and training dataset are publicly available on Hugging Face.
inproceedings TKF26
WMT 2026
11th Conference on Machine Translation. Budapest, Hungary, Oct 28-29, 2026. To be published. Preprint available.Authors
T. K. Temesgen • S. Kurtyigit • A. FraserLinks
URLResearch Area
BibTeXKey: TKF26