LaCoMSA: Language-Consistency Multilingual Self-Alignment with latent representation rewarding
Loading...
Date
2026-03-29
Authors
Tran, Khanh Tung
O’Sullivan, Barry
Nguyen, Hoang D.
Journal Title
Journal ISSN
Volume Title
Publisher
Association for Computational Linguistics (ACL)
Published Version
Abstract
Large Language Models (LLMs) have achieved impressive performance yet remain inconsistent across languages, often defaulting to high-resource outputs such as English. Existing multilingual alignment methods mitigate these issues through preference optimization but rely on external supervision, such as translation systems or English-biased signal. We propose Multilingual Self-Alignment (MSA), a targeted preference optimization framework that leverages an LLM’s own latent representations as intrinsic supervision signals, rewarding lower-resource language outputs based on their alignment with high-resource (English) counterparts in the “semantic hub”. We further introduce Language-Consistency MSA (LaCoMSA), which augments MSA with a final-layer language-consistency factor to prevent off-target generation. Integrated with Direct Preference Optimization, LaCoMSA improves a Llama 3 8B-based model multilingual win rates by up to 6.8% absolute (55.0% relatively) on X-AlpacaEval and achieves consistent gains across benchmarks and models. Our findings demonstrate that LaCoMSA can serve as an effective and scalable mechanism, opening a new venue toward multilingual self-alignment.
Description
© 2026, Association for Computational Linguistics.
Keywords
Multilingual Self-Alignment , Large Language Models (LLMs) , [ComputerScience] , Computational Theory and Mathematics , Software , Linguistics and Language
Citation
Tran, K T, O’Sullivan, B & Nguyen, H D 2026, LaCoMSA: Language-Consistency Multilingual Self-Alignment with latent representation rewarding. in V Demberg, K Inui & L Marquez Villodre (eds), Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers) : Rabat, Morocco, 24–29 March 2026. EACL 2026 - 19th Conference of the European Chapter of the Association for Computational Linguistics, Proceedings of the Conference, Vol. 1 - (Long Papers), vol. 1, Association for Computational Linguistics (ACL), pp. 4839-4853, 19th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2026, Rabat, Morocco, 24/03/26. https://doi.org/10.18653/v1/2026.eacl-long.224
conference
conference
Link to publisher’s version
Copyright
cc_by