Metraipolis › archive › dc_2194dc708276
Multilingual Safety Signals Are Multi-Layered: Filtering Safety-Degrading Data for Safer LLMs — arXiv:2609.22144
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.22144
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T06:48:54.972Z
- SHA-256 of the archived text
bed3f93457ffe46515ac23e603678f6063e69718f26e9d3ae6967cfa06a53f5f- Stable URL of this page
https://metraipolis.com/documento/dc_2194dc708276
How to cite
Multilingual Safety Signals Are Multi-Layered: Filtering Safety-Degrading Data for Safer LLMs — arXiv:2609.22144 — arXiv (en), https://arxiv.org/abs/2609.22144 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2194dc708276 (Metraipolis archive dc_2194dc708276, sha256 bed3f93457ffe46515ac23e603678f6063e69718f26e9d3ae6967cfa06a53f5f)
BibTeX
@misc{dc2194dc708276,
title = {Multilingual Safety Signals Are Multi-Layered: Filtering Safety-Degrading Data for Safer LLMs — arXiv:2609.22144},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.22144},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2194dc708276; sha256 bed3f93457ffe46515ac23e603678f6063e69718f26e9d3ae6967cfa06a53f5f},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.22144. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Preserving safety alignment during large language models fine-tuning is critical, however, recent studies have demonstrated that even benign fine-tuning data may contain safety-degrading samples that silently undermine safety alignment. Existing approaches typically identify such samples using representations from a single safety-sensitive layer. While this assumption has shown effectiveness in monolingual settings, its validity for multilingual models remains unclear due to potential cross-lingual differences in representation patterns. Through a cross-lingual analysis, we show that sensitive layers are only partially shared across languages, with safety-relevant signals often distributed across multiple layers. Motivated by these observations, we propose MMSAFE, a multi-layer framework for multilingual safety-degrading data identification that captures both shared and language-specific safety signals. Extensive experiments across multiple models, languages, and safety benchmarks demonstrate that MMSAFE reduces the average harmful-response ratio by 60% compared with random filtering and achieves stronger average performance than the strongest single-layer baseline, demonstrating the effectiveness of multi-layer modeling for robust multilingual safety alignment. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-22 in cs.CL). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.22144 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-09T06:48:44.844Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-22?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).