UCCIX: Irish-eXcellence Large Language Model
Open accessAbstract
The development of Large Language Models (LLMs) has predominantly focused on high-resource languages, leaving extremely low-resource languages like Irish with limited representation. This work presents UCCIX, a pioneering effort on the development of an open-source Irish-based LLM. We propose a novel framework for continued pre-training of LLMs specifically adapted for extremely low-resource languages, requiring only a fraction of the textual data typically needed for training LLMs according to scaling laws. Our model, based on Llama 2-13B [23], outperforms much larger models on Irish language tasks with up to 12% performance improvement, showcasing the effectiveness and efficiency of our approach. We also contribute comprehensive Irish benchmarking datasets, including IrishQA, a question-answering dataset, and Irish version of MT-bench [28]. These datasets enable rigorous evaluation and facilitate future research in Irish LLM systems. Our work aims to preserve and promote the Irish language, knowledge, and culture of Ireland in the digital era while providing a framework for adapting LLMs to other indigenous languages.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
UCCIX's creators explicitly frame it as a blueprint for other low-resource and Indigenous languages, with the dual goals of preserving Irish in the digital era and promoting Irish knowledge and culture globally.
“Its creators explicitly frame it as a blueprint for other low-resource and Indigenous languages, with the dual goals of preserving Irish in the digital era and promoting Irish knowledge and culture globally.”
From Data colonialism and indigenous languages in AI: a critical review of existing initiatives and their struggles with data sovereignty (2026) · cites this paper as Tran et al. 2024
The abstract says the work aims to preserve and promote the Irish language, knowledge, and culture in the digital era while providing a framework for other Indigenous languages. The Introduction and Conclusion also describe it as a blueprint for other low-resource languages. Our copy does not specify a goal of promoting Irish knowledge and culture globally. (LLM, conf=0.85)
Experimental — accuracy not yet measuredUCCIX is an open-source Irish LLM introduced in 2024.
“A landmark effort is UCCIX, an open-source Irish LLM introduced in 2024 (Tran et al. 2024).”
From Data colonialism and indigenous languages in AI: a critical review of existing initiatives and their struggles with data sovereignty (2026) · cites this paper as Tran et al. 2024
UCCIX contributed new evaluation benchmarks to enable rigorous testing of Irish language understanding and generation.
“In addition, UCCIX contributed new evaluation benchmarksfoot_0 to enable rigorous testing of Irish language understanding and generation.”
From Data colonialism and indigenous languages in AI: a critical review of existing initiatives and their struggles with data sovereignty (2026) · cites this paper as Tran et al. 2024
UCCIX, at 13B parameters, outperforms even much larger general models on Irish-specific tasks by up to 12%.
“The resulting model, at 13B parameters, outperforms even much larger general models on Irish-specific tasks by up to 12%.”
From Data colonialism and indigenous languages in AI: a critical review of existing initiatives and their struggles with data sovereignty (2026) · cites this paper as Tran et al. 2024
UCCIX achieved strong results 'with only a fraction of the textual data typically needed' according to standard scaling laws.
“Remarkably, they achieved strong results “with only a fraction of the textual data typically needed” according to standard scaling laws (Tran et al. 2024).”
From Data colonialism and indigenous languages in AI: a critical review of existing initiatives and their struggles with data sovereignty (2026) · cites this paper as Tran et al. 2024
The open-source release of UCCIX explicitly aims to enable adapting LLMs to other Indigenous languages using its framework.
“Indeed, the open-source release of UCCIX explicitly aims to enable”
From Data colonialism and indigenous languages in AI: a critical review of existing initiatives and their struggles with data sovereignty (2026) · cites this paper as Tran et al. 2024
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
No references could be resolved yet.
Research works printed but not linked: 18
- [2][2] David Adelani, Hannah Liu, Xiaoyu Shen, Nikita Vassilyev, Jesujoba Alabi, Yanke Mao, Haonan Gao, and En-Shiun Lee, 'SIB-200: A sim- ple, inclusive, and big evaluation dataset for topic classification in 200+ languages and dialects', in Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), eds., Yvette Graham and Matthew Purver, pp. 226–245, St. Julian's, Malta, (March 2024). Association for Com- putational Linguistics.
- [3][3] Zhangir Azerbayev, Hailey Schoelkopf, Keiran Paster, Marco Dos San- tos, Stephen Marcus McAleer, Albert Q. Jiang, Jia Deng, Stella Bider- man, and Sean Welleck, 'Llemma: An open language model for math- ematics', in The Twelfth International Conference on Learning Repre- sentations, (2024).
- [4][4] Marta Bañón, Pinzhen Chen, Barry Haddow, Kenneth Heafield, Hieu Hoang, Miquel Esplà-Gomis, Mikel L. Forcada, Amir Kamran, Faheem Kirefu, Philipp Koehn, Sergio Ortiz Rojas, Leopoldo Pla Sempere, Gema Ramírez-Sánchez, Elsa Sarrías, Marek Strelec, Brian Thompson, William Waites, Dion Wiggins, and Jaume Zaragoza, 'ParaCrawl: Web- scale acquisition of parallel corpora', in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, eds., Dan Ju- rafsky, Joyce Chai, Natalie Schluter, and Joel Tetreault, pp. 4555–4567, Online, (July 2020). Association for Computational Linguistics.
- [5][5] James Barry, Joachim Wagner, Lauren Cassidy, Alan Cowap, Teresa Lynn, Abigail Walsh, Mícheál J. Ó Meachair, and Jennifer Foster, 'gaBERT — an Irish language model', in Proceedings of the Thirteenth Language Resources and Evaluation Conference, eds., Nicoletta Cal- zolari, Frédéric Béchet, Philippe Blache, Khalid Choukri, Christopher Cieri, Thierry Declerck, Sara Goggi, Hitoshi Isahara, Bente Maegaard, Joseph Mariani, Hélène Mazo, Jan Odijk, and Stelios Piperidis, pp. 4774–4788, Marseille, France, (June 2022). European Language Re- sources Association.
- [6][6] Yupeng Chang, Xu Wang, Jindong Wang, Yuan Wu, Linyi Yang, Kaijie Zhu, Hao Chen, Xiaoyuan Yi, Cunxiang Wang, Yidong Wang, Wei Ye, Yue Zhang, Yi Chang, Philip S. Yu, Qiang Yang, and Xing Xie, 'A sur- vey on evaluation of large language models', ACM Trans. Intell. Syst. Technol., 15(3), (mar 2024).
- [9][9] Dirk Goldhahn, Thomas Eckart, and Uwe Quasthoff, 'Building large monolingual dictionaries at the Leipzig corpora collection: From 100 to 200 languages', in Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC'12), eds., Nicoletta Calzolari, Khalid Choukri, Thierry Declerck, Mehmet Uğur Doğan, Bente Maegaard, Joseph Mariani, Asuncion Moreno, Jan Odijk, and Stelios Piperidis, pp. 759–765, Istanbul, Turkey, (May 2012). European Language Resources Association (ELRA).
- [11][11] Jordan Hoffmann et al., 'An empirical analysis of compute-optimal large language model training', in Advances in Neural Information Pro- cessing Systems, eds., Alice H. Oh, Alekh Agarwal, Danielle Belgrave, and Kyunghyun Cho, (2022).
- [12][12] Ayyoob ImaniGooghari, Peiqin Lin, Amir Hossein Kargaran, Sil- via Severini, Masoud Jalili Sabet, Nora Kassner, Chunlan Ma, Hel- mut Schmid, André Martins, François Yvon, and Hinrich Schütze, 'Glot500: Scaling multilingual corpora and language models to 500 languages', in Proceedings of the 61st Annual Meeting of the Asso- ciation for Computational Linguistics (Volume 1: Long Papers), eds., Anna Rogers, Jordan Boyd-Graber, and Naoaki Okazaki, pp. 1082– 1117, Toronto, Canada, (July 2023). Association for Computational Linguistics.
- [13][13] Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bam- ford, Devendra Singh Chaplot, Diego de Las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix, and William El Sayed, 'Mis- tral 7b', CoRR, abs/2310.06825, (2023). · arXiv
- [14][14] Seamus Lankford, Haithem Afli, and Andy Way, 'Machine translation in the covid domain: an English-Irish case study for LoResMT 2021', in Proceedings of the 4th Workshop on Technologies for MT of Low Resource Languages (LoResMT2021), eds., John Ortega, Atul Kr. Ojha, Katharina Kann, and Chao-Hong Liu, pp. 144–150, Virtual, (August 2021). Association for Machine Translation in the Americas.
- [15][15] Risto Luukkonen et al., 'FinGPT: Large generative models for a small language', in Proceedings of the 2023 Conference on Empirical Meth- ods in Natural Language Processing, eds., Houda Bouamor, Juan Pino, and Kalika Bali, pp. 2710–2726, Singapore, (December 2023). Associ- ation for Computational Linguistics.
- [17][17] Simon Mille, Elaine Uí Dhonnchadha, Lauren Cassidy, Brian Davis, Stamatia Dasiopoulou, and Anya Belz, 'Generating Irish text with a flexible plug-and-play architecture', in Proceedings of the 2nd Work- shop on Pattern-based Approaches to NLP in the Age of Deep Learn- ing, eds., Mihai Surdeanu, Ellen Riloff, Laura Chiticariu, Dayne Fri- etag, Gus Hahn-Powell, Clayton T. Morrison, Enrique Noriega-Atala, Rebecca Sharp, and Marco Valenzuela-Escarcega, pp. 25–42, Singa- pore, (December 2023). Association for Computational Linguistics.
- [18][18] Swaroop Mishra, Daniel Khashabi, Chitta Baral, and Hannaneh Ha- jishirzi, 'Cross-task generalization via natural language crowdsourcing instructions', in Proceedings of the 60th Annual Meeting of the Asso- ciation for Computational Linguistics (Volume 1: Long Papers), eds., Smaranda Muresan, Preslav Nakov, and Aline Villavicencio, pp. 3470– 3487, Dublin, Ireland, (May 2022). Association for Computational Lin- guistics.
- [19][19] Niklas Muennighoff, Alexander M Rush, Boaz Barak, Teven Le Scao, Nouamane Tazi, Aleksandra Piktus, Sampo Pyysalo, Thomas Wolf, and Colin Raffel, 'Scaling data-constrained language models', in Thirty- seventh Conference on Neural Information Processing Systems, (2023).
- [21][21] Long Ouyang et al., 'Training language models to follow instructions with human feedback', in Advances in Neural Information Processing Systems, eds., Alice H. Oh, Alekh Agarwal, Danielle Belgrave, and Kyunghyun Cho, (2022).
- [25][25] BigScience Workshop et al. Bloom: A 176b-parameter open-access multilingual language model, 2022.
- [27][27] Zheng Xin Yong et al., 'BLOOM+1: Adding language support to BLOOM for zero-shot prompting', in Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), eds., Anna Rogers, Jordan Boyd-Graber, and Naoaki Okazaki, pp. 11682–11703, Toronto, Canada, (July 2023). Association for Computational Linguistics.
- [28][28] Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhang- hao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica, 'Judging LLM-as-a- judge with MT-bench and chatbot arena', in Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track, (2023). 4 https://cloud.google.com/translate?hl=en https://cloud.google.com/translate?hl=en Introduction and Contribution UCCIX: Irish-eXcellence Large Language Model Benchmarking Results Conclusion
Cited by
Sorted by citations
Multi-Agent Collaboration Mechanisms: A Survey of LLMs
Tran, K.-T., Dao, D., Nguyen, M.-D., Pham, Q.-V., O'Sullivan, B., & Nguyen, H. D. (2025) · arXiv · doi:10.48550/arxiv.2501.06322 · Cited by 26 · see citing works
Self-citationKwok, J. C. Y. (2026) · AI & SOCIETY · doi:10.1007/s00146-026-03091-w · Cited by 4 · see citing works
- Authors
- Tran, K.-T., O’Sullivan, B., & Nguyen, H. D.
- Affiliations
- University College Cork (IE)
- Venue
- Frontiers in Artificial Intelligence and Applications
- Publisher
- IOS Press
- Published
- 2024-10-16
- ISSN
- 0922-6389 (ISSN-L)
- 1879-8314
- OpenAlex
- W4403479290
- Cited by
- 2
- References
- 0
- Citations by year
- 2026: 1 · 2025: 1
- FWCI
- 1.31
- Type
- book chapter
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- Frontiers in artificial intelligence and applications (publishedVersion, cc-by-nc)
- arXiv (Cornell University) (submittedVersion, cc-by)
- Cork Open Research Archive (University College Cork, Ireland) (submittedVersion, cc-by-nc)
- Cork Open Research Archive (University College Cork, Ireland) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- Nvidia
- Science Foundation Ireland (18/CRT/6223, 12/RC/2289_P2, 12/RC/, 12/RC/2289)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- Irish language, large language models, continual pre-training, open-source language models, question answering dataset, MT-Bench, language preservation, indigenous languages
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
