Translation Asymmetry in LLMs as a Data Augmentation Factor: A Case Study for 6 Romansh Language Varieties
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914425151160320 |
|---|---|
| author | Vamvas, Jannis Prat, Ignacio Pérez Heldstab, Angela Fischer, Dominic P. Ahmadi, Sina Sennrich, Rico |
| author_facet | Vamvas, Jannis Prat, Ignacio Pérez Heldstab, Angela Fischer, Dominic P. Ahmadi, Sina Sennrich, Rico |
| contents | Recent strategies for low-resource machine translation rely on LLMs to generate synthetic data from higher-resource languages. We find that this method fails for Romansh, because LLMs tend to confuse its 6 distinct language varieties. Our experiments show that instead, the direction of data augmentation should be aligned with the resource gradient between source and target language. This approach surpasses Gemini 3 Pro in the lowest-resource variety of Romansh by 23 BLEU. A human evaluation confirms that our experiments yield the first model that generates fluent translations in the individual Romansh varieties. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_25489 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Translation Asymmetry in LLMs as a Data Augmentation Factor: A Case Study for 6 Romansh Language Varieties Vamvas, Jannis Prat, Ignacio Pérez Heldstab, Angela Fischer, Dominic P. Ahmadi, Sina Sennrich, Rico Computation and Language Recent strategies for low-resource machine translation rely on LLMs to generate synthetic data from higher-resource languages. We find that this method fails for Romansh, because LLMs tend to confuse its 6 distinct language varieties. Our experiments show that instead, the direction of data augmentation should be aligned with the resource gradient between source and target language. This approach surpasses Gemini 3 Pro in the lowest-resource variety of Romansh by 23 BLEU. A human evaluation confirms that our experiments yield the first model that generates fluent translations in the individual Romansh varieties. |
| title | Translation Asymmetry in LLMs as a Data Augmentation Factor: A Case Study for 6 Romansh Language Varieties |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.25489 |