Tiny Aya: Bridging Scale and Multilingual Depth
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866917334706290688 |
|---|---|
| author | Salamanca, Alejandro R. Abagyan, Diana D'souza, Daniel Khairi, Ammar Mora, David Dash, Saurabh Aryabumi, Viraat Rajaee, Sara Mofakhami, Mehrnaz Sahu, Ananya Euyang, Thomas Prince, Brittawnya Smith, Madeline Lin, Hangyu Locatelli, Acyr Hooker, Sara Kocmi, Tom Gomez, Aidan Zhang, Ivan Blunsom, Phil Frosst, Nick Pineau, Joelle Ermis, Beyza Üstün, Ahmet Kreutzer, Julia Fadaee, Marzieh |
| author_facet | Salamanca, Alejandro R. Abagyan, Diana D'souza, Daniel Khairi, Ammar Mora, David Dash, Saurabh Aryabumi, Viraat Rajaee, Sara Mofakhami, Mehrnaz Sahu, Ananya Euyang, Thomas Prince, Brittawnya Smith, Madeline Lin, Hangyu Locatelli, Acyr Hooker, Sara Kocmi, Tom Gomez, Aidan Zhang, Ivan Blunsom, Phil Frosst, Nick Pineau, Joelle Ermis, Beyza Üstün, Ahmet Kreutzer, Julia Fadaee, Marzieh |
| contents | Tiny Aya redefines what a small multilingual language model can achieve. Trained on 70 languages and refined through region-aware posttraining, it delivers state-of-the-art in translation quality, strong multilingual understanding, and high-quality target-language generation, all with just 3.35B parameters. The release includes a pretrained foundation model, a globally balanced instruction-tuned variant, and three region-specialized models targeting languages from Africa, South Asia, Europe, Asia-Pacific, and West Asia. This report details the training strategy, data composition, and comprehensive evaluation framework behind Tiny Aya, and presents an alternative scaling path for multilingual AI: one centered on efficiency, balanced performance across languages, and practical deployment. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_11510 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Tiny Aya: Bridging Scale and Multilingual Depth Salamanca, Alejandro R. Abagyan, Diana D'souza, Daniel Khairi, Ammar Mora, David Dash, Saurabh Aryabumi, Viraat Rajaee, Sara Mofakhami, Mehrnaz Sahu, Ananya Euyang, Thomas Prince, Brittawnya Smith, Madeline Lin, Hangyu Locatelli, Acyr Hooker, Sara Kocmi, Tom Gomez, Aidan Zhang, Ivan Blunsom, Phil Frosst, Nick Pineau, Joelle Ermis, Beyza Üstün, Ahmet Kreutzer, Julia Fadaee, Marzieh Computation and Language Tiny Aya redefines what a small multilingual language model can achieve. Trained on 70 languages and refined through region-aware posttraining, it delivers state-of-the-art in translation quality, strong multilingual understanding, and high-quality target-language generation, all with just 3.35B parameters. The release includes a pretrained foundation model, a globally balanced instruction-tuned variant, and three region-specialized models targeting languages from Africa, South Asia, Europe, Asia-Pacific, and West Asia. This report details the training strategy, data composition, and comprehensive evaluation framework behind Tiny Aya, and presents an alternative scaling path for multilingual AI: one centered on efficiency, balanced performance across languages, and practical deployment. |
| title | Tiny Aya: Bridging Scale and Multilingual Depth |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.11510 |