---
title: "Layer-wise Swapping for Generalizable Multilingual Safety (2026)"
description: "Shin and Hwang propose a fix that needs no training. 39 references, cited by 1 paper here."
authors: ["Hyunseo Shin", "Wonseok Hwang"]
year: 2026
venue: "Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)"
source: "https://aclanthology.org/2026.eacl-long.98/"
doi: "10.18653/v1/2026.eacl-long.98"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/shin-2026-layer-wise-swapping-generalizable-multilingual/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/shin-2026-layer-wise-swapping-generalizable-multilingual/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/14/) / Layer-wise Swapping for Generalizable Multilingual Safety

# Layer-wise Swapping for Generalizable Multilingual Safety

**Authors:** Hyunseo Shin and [Wonseok Hwang](https://md.research.elamur.ai/free/en/authors/wonseok-hwang/) · **Year:** 2026 · **Venue:** [Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/eacl/) ([EACL 2026](https://md.research.elamur.ai/free/en/venues/eacl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.eacl-long.98/) · [PDF](https://aclanthology.org/2026.eacl-long.98.pdf) · [DOI 10.18653/v1/2026.eacl-long.98](https://doi.org/10.18653/v1/2026.eacl-long.98) · [OpenAlex](https://openalex.org/W7140074690)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · 1 in our library, of the 149,439 papers whose bibliographies we read

## Summary
Safety tuning for LLMs relies mostly on English data. As a result, models adapted to low-resource languages through instruction fine-tuning often become noticeably less safe. Shin and Hwang propose a fix that needs no training. Starting from a shared base model, they treat an English safety expert and a language expert as two task vectors and recombine their parameters.

Earlier layer swapping used manually chosen layers. This method instead scores every attention and MLP module by how strongly each expert changed it relative to the base, using normalized update ratios. If one expert clearly dominates a module, that module is copied from it. Modules with similar scores are averaged (threshold τ = 0.001, weight α = 0.5).

The authors tested the method on Korean, Bengali, Swahili and Telugu with LLaMA 3.1 8B Instruct and Qwen3 8B. The merged models keep general scores on MMMLU, BELEBELE and MGSM close to the language experts while producing fewer harmful answers on MultiJail. This is relevant for any language where dedicated safety data is scarce, Ukrainian included, although Ukrainian was not evaluated.

## Key points
- **Training setup:** Language experts were fine-tuned with torchtune on 70k–80k instruction samples per language. The safety expert was tuned on only 2k English safety instructions (Bianchi et al., 2024). MultiJail prompts for Telugu and Korean were translated with GPT-4o.
- **Module analysis (Bengali):** Attention and MLP modules show opposite patterns of expert dominance across layer depth. This suggests that safety and language skills sit in different components, which motivates choosing at the module level rather than the whole-layer level.
- **Ablations on Bengali:** τ = 0.001 beat τ = 0 on the general benchmarks. α = 0.5 gave the best overall trade-off, with MultiJail unsafe rates of 22.3 for LLaMA and 14.6 for Qwen.
- **Judge reliability:** On 50 harmful prompts per language, Qwen Guard agreed with human labels 85.5% of the time on average, versus 77.0% for Gemma Guard. Both judges were weakest on Swahili (70% and 64%).
- **Limitations and code:** The authors note two limitations: results depend on LLM-based safety judges, and the merge is static rather than input-dependent. Code is available at github.com/00HS/layer-wise_swapping.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [University of Seoul](https://md.research.elamur.ai/free/en/universities/university-of-seoul/) — University, Seoul, South Korea · 4 works · 284 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (2): [🌍 Multilingual & cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · 828 works · [🛡️ Bias, fairness & safety](https://md.research.elamur.ai/free/en/topics/bias-fairness-safety/) · 517 works

## Cite

Cite as: Hyunseo Shin and Wonseok Hwang. 2026. Layer-wise Swapping for Generalizable Multilingual Safety. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{shin2026layer,
  title = {{Layer-wise Swapping for Generalizable Multilingual Safety}},
  author = {Hyunseo Shin and Wonseok Hwang},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-long.98},
  url = {https://doi.org/10.18653/v1/2026.eacl-long.98}
}
```

- [← previous LATE-IIMAS at Semeval-2026 Task 13: Evaluating GNNs, PLMs, LLMs, and Stylometry for Automatic Code Identification](https://md.research.elamur.ai/free/en/papers/valdez-2026-late-iimas-semeval-2026-task/)
- [next → Learning from Contrasts: Synthesizing Reasoning Paths from Diverse Search Trajectories](https://md.research.elamur.ai/free/en/papers/liu-2026-learning-contrasts-synthesizing-reasoning-paths/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/14/)

## References

39 references, 18 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. AI@Meta. 2024. Llama 3 model card. [find](https://scholar.google.com/scholar?q=AI%40Meta%2E+2024%2E+Llama+3+model+card%2E)
2. Anton Alexandrov, Veselin Raychev, Mark Niklas Müller, Ce Zhang, Martin Vechev, and Kristina Toutanova. 2024. Mitigating catastrophic forgetting in language transfer via model merging. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 17167–17186, Miami, Florida, USA. A… [find](https://scholar.google.com/scholar?q=Anton+Alexandrov%2C+Veselin+Raychev%2C+Mark+Niklas+M%C3%BCller%2C+Ce+Zhang%2C+Martin+Vechev%2C+and+Kristina+Toutanova%2E+2024%2E+Mitigating+catastrophic+forgetting+in+language+transfer+via+model+merging%2E+In+Findings+of+)
3. Lucas Bandarkar, Davis Liang, Benjamin Muller, Mikel Artetxe, Satya Narayan Shukla, Donald Husa, Naman Goyal, Abhinandan Krishnan, Luke Zettlemoyer, and Madian Khabsa. 2024. [The belebele benchmark: a parallel reading comprehension dataset in 122 language variants](https://md.research.elamur.ai/free/en/papers/bandarkar-2023-belebele-benchmark-parallel-reading-comprehension/). In Proceedings of the 62nd Annual M…
4. Lucas Bandarkar, Benjamin Muller, Pritish Yuvraj, Rui Hou, Nayan Singhal, Hongjiang Lv, and Bing Liu. 2025. Layer swapping for zero-shot cross-lingual transfer in large language models. In The Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025. O… [find](https://scholar.google.com/scholar?q=Lucas+Bandarkar%2C+Benjamin+Muller%2C+Pritish+Yuvraj%2C+Rui+Hou%2C+Nayan+Singhal%2C+Hongjiang+Lv%2C+and+Bing+Liu%2E+2025%2E+Layer+swapping+for+zero%2Dshot+cross%2Dlingual+transfer+in+large+language+models%2E+In+The+Thirtee)
5. Federico Bianchi, Mirac Suzgun, Giuseppe Attanasio, Paul Rottger, Dan Jurafsky, Tatsunori Hashimoto, and James Zou. 2024. [Safety-tuned LLaMAs: Lessons from improving the safety of large language models that follow instructions](https://md.research.elamur.ai/free/en/papers/bianchi-2023-safety-tuned-llamas-lessons-improving/). In The Twelfth International Conference on Learning Representations.
6. Nadezhda Chirkova and Vassilina Nikoulina. 2024. Zero-shot cross-lingual transfer in instruction tuning of large language models. In Proceedings of the 17th International Natural Language Generation Conference, pages 695–708, Tokyo, Japan. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Nadezhda+Chirkova+and+Vassilina+Nikoulina%2E+2024%2E+Zero%2Dshot+cross%2Dlingual+transfer+in+instruction+tuning+of+large+language+models%2E+In+Proceedings+of+the+17th+International+Natural+Language+Generation+C)
7. Alexandra Chronopoulou, Jonas Pfeiffer, Joshua Maynez, Xinyi Wang, Sebastian Ruder, and Priyanka Agrawal. 2024. Language and task arithmetic with parameter-efficient layers for zero-shot summarization. In Proceedings of the Fourth Workshop on Multilingual Representation Learning (MRL 2024), pages 11… [find](https://scholar.google.com/scholar?q=Alexandra+Chronopoulou%2C+Jonas+Pfeiffer%2C+Joshua+Maynez%2C+Xinyi+Wang%2C+Sebastian+Ruder%2C+and+Priyanka+Agrawal%2E+2024%2E+Language+and+task+arithmetic+with+parameter%2Defficient+layers+for+zero%2Dshot+summarization)
8. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. [Training verifiers to solve math word problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/). arXiv preprint arXiv:2110.14168.
9. Rui Dai, Sile Hu, Xu Shen, Yonggang Zhang, Xinmei Tian, and Jieping Ye. 2025. Leveraging submodule linearity enhances task arithmetic performance in LLMs. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Rui+Dai%2C+Sile+Hu%2C+Xu+Shen%2C+Yonggang+Zhang%2C+Xinmei+Tian%2C+and+Jieping+Ye%2E+2025%2E+Leveraging+submodule+linearity+enhances+task+arithmetic+performance+in+LLMs%2E+In+The+Thirteenth+International+Conference+on)
10. Yue Deng, Wenxuan Zhang, Sinno Jialin Pan, and Lidong Bing. 2024. [Multilingual jailbreak challenges in large language models](https://md.research.elamur.ai/free/en/papers/deng-2023-multilingual-jailbreak-challenges-large-language/). In The Twelfth International Conference on Learning Representations.
11. Jiwoo Hong, Noah Lee, Rodrigo Martínez-Castaño, César Rodríguez, and James Thorne. 2025. Cross-lingual transfer of reward models in multilingual alignment. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Te… [find](https://scholar.google.com/scholar?q=Jiwoo+Hong%2C+Noah+Lee%2C+Rodrigo+Mart%C3%ADnez%2DCasta%C3%B1o%2C+C%C3%A9sar+Rodr%C3%ADguez%2C+and+James+Thorne%2E+2025%2E+Cross%2Dlingual+transfer+of+reward+models+in+multilingual+alignment%2E+In+Proceedings+of+the+2025+Conference+of+the)
12. Kaiyu Huang, Peng Li, Jin Ma, Ting Yao, and Yang Liu. 2023. Knowledge transfer in incremental learning for multilingual neural machine translation. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15286–15304, Toronto, Canada.… [find](https://scholar.google.com/scholar?q=Kaiyu+Huang%2C+Peng+Li%2C+Jin+Ma%2C+Ting+Yao%2C+and+Yang+Liu%2E+2023%2E+Knowledge+transfer+in+incremental+learning+for+multilingual+neural+machine+translation%2E+In+Proceedings+of+the+61st+Annual+Meeting+of+the+Ass)
13. Gabriel Ilharco, Marco Tulio Ribeiro, Mitchell Wortsman, Ludwig Schmidt, Hannaneh Hajishirzi, and Ali Farhadi. 2023. [Editing models with task arithmetic](https://md.research.elamur.ai/free/en/papers/ilharco-2022-editing-models-task-arithmetic/). In The Eleventh International Conference on Learning Representations.
14. Mohammed Safi Ur Rahman Khan, Priyam Mehta, Ananth Sankar, Umashankar Kumaravelan, Sumanth Doddapaneni, Suriyaprasaad G, Varun Balan G, Sparsh Jain, Anoop Kunchukuttan, Pratyush Kumar, Raj Dabre, and Mitesh M. Khapra. 2024. [Indicllmsuite: A blueprint for creating pre-training and fine-tuning dataset…](https://md.research.elamur.ai/free/en/papers/khan-2024-indicllmsuite-blueprint-creating-pre-training/)
15. Takeshi Kojima, Itsuki Okimura, Yusuke Iwasawa, Hitomi Yanaka, and Yutaka Matsuo. 2024. [On the multilingual ability of decoder-based pre-trained language models: Finding and controlling language-specific neurons](https://md.research.elamur.ai/free/en/papers/kojima-2024-multilingual-ability-decoder-based-pre/). In Proceedings of the 2024 Conference of the North American Chapter of the Association…
16. Shen Li, Liuyi Yao, Lan Zhang, and Yaliang Li. 2025. Safety layers in aligned large language models: The key to LLM security. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Shen+Li%2C+Liuyi+Yao%2C+Lan+Zhang%2C+and+Yaliang+Li%2E+2025%2E+Safety+layers+in+aligned+large+language+models%3A+The+key+to+LLM+security%2E+In+The+Thirteenth+International+Conference+on+Learning+Representations%2E)
17. AI @ Meta Llama Team. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). Preprint, arXiv:2407.21783.
18. Niklas Muennighoff, Thomas Wang, Lintang Sutawika, Adam Roberts, Stella Biderman, Teven Le Scao, M Saiful Bari, Sheng Shen, Zheng-Xin Yong, Hailey Schoelkopf, Xiangru Tang, Dragomir Radev, Alham Fikri Aji, Khalid Almubarak, Samuel Albanie, Zaid Alyafeai, Albert Webson, Edward Raff, and Colin Raffel.… — [Crosslingual Generalization through Multitask Finetuning](https://md.research.elamur.ai/free/en/papers/muennighoff-2022-crosslingual-generalization-through-multitask-finetuning/)
19. OpenAI, :, Aaron Hurst, Adam Lerer, Adam P. Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, Aleksander M adry, Alex Baker-Whitcomb, Alex Beutel, Alex Borzunov, Alex Carney, Alex Chow, Alex Kirillov, and 401 others. 2024. [Gpt-4o system card](https://md.research.elamur.ai/free/en/papers/openai-2024-gpt-4o-system-card/).…
20. OpenAI. 2024. Multilingual massive multitask language understanding (mmmlu). [https://huggingface.co/datasets/openai/MMMLU](https://huggingface.co/datasets/openai/MMMLU). [↗](https://huggingface.co/datasets/openai/MMMLU)
21. Lingfeng Shen, Weiting Tan, Sihao Chen, Yunmo Chen, Jingyu Zhang, Haoran Xu, Boyuan Zheng, Philipp Koehn, and Daniel Khashabi. 2024. [The language barrier: Dissecting safety challenges of LLMs in multilingual contexts](https://md.research.elamur.ai/free/en/papers/shen-2024-language-barrier-dissecting-safety-challenges/). In Findings of the Association for Computational Linguistics: ACL 2024, pages 2668…
22. Freda Shi, Mirac Suzgun, Markus Freitag, Xuezhi Wang, Suraj Srivats, Soroush Vosoughi, Hyung Won Chung, Yi Tay, Sebastian Ruder, Denny Zhou, Dipanjan Das, and Jason Wei. 2023. [Language models are multilingual chain-of-thought reasoners](https://md.research.elamur.ai/free/en/papers/shi-2022-language-models-multilingual-chain-thought/). In The Eleventh International Conference on Learning Representa…
23. Enneng Yang, Li Shen, Zhenyi Wang, Guibing Guo, Xiaojun Chen, Xingwei Wang, and Dacheng Tao. 2024. Representation surgery for multi-task model merging. In Proceedings of the 41st International Conference on Machine Learning, volume 235 of Proceedings of Machine Learning Research, pages 56332–56356.… [find](https://scholar.google.com/scholar?q=Enneng+Yang%2C+Li+Shen%2C+Zhenyi+Wang%2C+Guibing+Guo%2C+Xiaojun+Chen%2C+Xingwei+Wang%2C+and+Dacheng+Tao%2E+2024%2E+Representation+surgery+for+multi%2Dtask+model+merging%2E+In+Proceedings+of+the+41st+International+Confere)
24. Shivalika Singh, Freddie Vargus, Daniel Dsouza, Börje F. Karlsson, Abinaya Mahendiran, Wei-Yin Ko, Herumb Shandilya, Jay Patel, Deividas Mataciunas, Laura OMahony, Mike Zhang, Ramith Hettiarachchi, Joseph Wilson, Marina Machado, Luisa Souza Moura, Dominik Krzemiński, Hakimeh Fadaei, Irem Ergün, Ife… — [Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning](https://md.research.elamur.ai/free/en/papers/singh-2024-aya-dataset-open-access-collection/)
25. Le Yu, Bowen Yu, Haiyang Yu, Fei Huang, and Yongbin Li. 2024. [Language models are super mario: Absorbing abilities from homologous models as a free lunch](https://md.research.elamur.ai/free/en/papers/yu-2023-language-models-super-mario-absorbing/). In Forty-first International Conference on Machine Learning.
26. Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. 2023. Stanford alpaca: An instruction-following llama model. [https://github.com/tatsu-lab/stanford_alpaca](https://github.com/tatsu-lab/stanford_alpaca). [↗](https://github.com/tatsu-lab/stanford_alpaca)
27. Gemma Team. 2025a. Gemma 3. [find](https://scholar.google.com/scholar?q=Gemma+Team%2E+2025a%2E+Gemma+3%2E)
28. NLLB Team, Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Heffernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, Anna Sun, Skyler Wang, Guillaume Wenzek, Al Youngblood, Bapi Akula, Loic Barrault, Gabriel Mejia Gonzalez, Prangthip Hansanti, and 20 o… [find](https://scholar.google.com/scholar?q=NLLB+Team%2C+Marta+R%2E+Costa%2Djuss%C3%A0%2C+James+Cross%2C+Onur+%C3%87elebi%2C+Maha+Elbayad%2C+Kenneth+Heafield%2C+Kevin+Heffernan%2C+Elahe+Kalbassi%2C+Janice+Lam%2C+Daniel+Licht%2C+Jean+Maillard%2C+Anna+Sun%2C+Skyler+Wang%2C+Guillaume+We)
29. Qwen Team. 2025b. [Qwen3 technical report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). Preprint, arXiv:2505.09388.
30. Qwen Team. 2025c. Qwen3guard technical report. [find](https://scholar.google.com/scholar?q=Qwen+Team%2E+2025c%2E+Qwen3guard+technical+report%2E)
31. Yuanchi Zhang, Yile Wang, Zijun Liu, Shuo Wang, Xiaolong Wang, Peng Li, Maosong Sun, and Yang Liu. 2024. [Enhancing multilingual capabilities of large language models through self-distillation from resource-rich languages](https://md.research.elamur.ai/free/en/papers/zhang-2024-enhancing-multilingual-capabilities-large-language/). In Proceedings of the 62nd Annual Meeting of the Association for Computational…
32. Wei Zhao, Zhe Li, Yige Li, Ye Zhang, and Jun Sun. 2024a. Defending large language models against jailbreak attacks via layer-specific editing. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 5094–5109, Miami, Florida, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Wei+Zhao%2C+Zhe+Li%2C+Yige+Li%2C+Ye+Zhang%2C+and+Jun+Sun%2E+2024a%2E+Defending+large+language+models+against+jailbreak+attacks+via+layer%2Dspecific+editing%2E+In+Findings+of+the+Association+for+Computational+Linguist)
33. Yiran Zhao, Wenxuan Zhang, Guizhen Chen, Kenji Kawaguchi, and Lidong Bing. 2024b. [How do large language models handle multilingualism?](https://md.research.elamur.ai/free/en/papers/zhao-2024-large-language-models-handle-multilingualism/) In The Thirty-eighth Annual Conference on Neural Information Processing Systems.
34. Yiran Zhao, Wenxuan Zhang, Yuxi Xie, Anirudh Goyal, Kenji Kawaguchi, and Michael Shieh. 2025. Understanding and enhancing safety mechanisms of LLMs via safety-specific neuron. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Yiran+Zhao%2C+Wenxuan+Zhang%2C+Yuxi+Xie%2C+Anirudh+Goyal%2C+Kenji+Kawaguchi%2C+and+Michael+Shieh%2E+2025%2E+Understanding+and+enhancing+safety+mechanisms+of+LLMs+via+safety%2Dspecific+neuron%2E+In+The+Thirteenth+Intern)
35. Atnafu Lambebo Tonja, Bonaventure FP Dossou, Jessica Ojo, Jenalea Rajab, Fadel Thior, Eric Peter Wairagala, Aremu Anuoluwapo, Pelonomi Moiloa, Jade Abbott, Vukosi Marivate, and 1 others. 2024. Inkubalm: A small language model for low-resource african languages. arXiv preprint arXiv:2408.17024. torch… — [arXiv:2408.17024](https://arxiv.org/abs/2408.17024) [find](https://scholar.google.com/scholar?q=Atnafu+Lambebo+Tonja%2C+Bonaventure+FP+Dossou%2C+Jessica+Ojo%2C+Jenalea+Rajab%2C+Fadel+Thior%2C+Eric+Peter+Wairagala%2C+Aremu+Anuoluwapo%2C+Pelonomi+Moiloa%2C+Jade+Abbott%2C+Vukosi+Marivate%2C+and+1+others%2E+2024%2E+Inkubal)
36. Wenxuan Wang, Zhaopeng Tu, Chang Chen, Youliang Yuan, Jen-tse Huang, Wenxiang Jiao, and Michael Lyu. 2024. All languages matter: On the multilingual safety of LLMs. In Findings of the Association for Computational Linguistics: ACL 2024, pages 5865– 5877, Bangkok, Thailand. Association for Computatio… [find](https://scholar.google.com/scholar?q=Wenxuan+Wang%2C+Zhaopeng+Tu%2C+Chang+Chen%2C+Youliang+Yuan%2C+Jen%2Dtse+Huang%2C+Wenxiang+Jiao%2C+and+Michael+Lyu%2E+2024%2E+All+languages+matter%3A+On+the+multilingual+safety+of+LLMs%2E+In+Findings+of+the+Association+for+)
37. Prateek Yadav, Derek Tam, Leshem Choshen, Colin A Raffel, and Mohit Bansal. 2023. [Ties-merging: Resolving interference when merging models](https://md.research.elamur.ai/free/en/papers/yadav-2023-ties-merging-resolving-interference-merging/). Advances in Neural Information Processing Systems, 36:7093–7115.
38. Type Name SFT Datasets URL KoAlpaca built from (Taori et al., 2023) Databricks-Dolly-15k-ko Aya (Korean subset) (Singh et al., 2024) NLLB (Korean subset) (Team et al., 2022) Aya (Bengali subset)(Singh et al., 2024) Indic-Align (Khan et al., 2024) Bengali BongChat-v1-253k NLLB (Bengali subset) (Team… [find](https://scholar.google.com/scholar?q=Type+Name+SFT+Datasets+URL+KoAlpaca+built+from+%28Taori+et+al%2E%2C+2023%29+Databricks%2DDolly%2D15k%2Dko+Aya+%28Korean+subset%29+%28Singh+et+al%2E%2C+2024%29+NLLB+%28Korean+subset%29+%28Team+et+al%2E%2C+2022%29+Aya+%28Bengali+subset%29%28Singh)
39. Apache 2.0 CC-BY-SA 3.0 Apache 2.0 ODC-By Apache 2.0 CC-BY 4.0 MIT ODC-By Apache 2.0 Apache 2.0 CC BY-NC 4.0 Apache 2.0 CC BY 4.0 MIT ODC-By MIT [https://huggingface.co/datasets/juletxara/mgsm](https://huggingface.co/datasets/juletxara/mgsm) https://huggingface.co/datasets/facebook/belebele https://huggingface.co/datasets/juletxara/mgsm https://hug… [↗](https://huggingface.co/datasets/juletxara/mgsm)

*Updated 31 October 2026*