Multilingual Amnesia: On the Transferability of Unlearning in Multilingual LLMs
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914243108929536 |
|---|---|
| author | Farashah, Alireza Dehghanpour Khandelwal, Aditi Fauchard, Marylou Shi, Zhuan Rostamzadeh, Negar Farnadi, Golnoosh |
| author_facet | Farashah, Alireza Dehghanpour Khandelwal, Aditi Fauchard, Marylou Shi, Zhuan Rostamzadeh, Negar Farnadi, Golnoosh |
| contents | As multilingual large language models become more widely used, ensuring their safety and fairness across diverse linguistic contexts presents unique challenges. While existing research on machine unlearning has primarily focused on monolingual settings, typically English, multilingual environments introduce additional complexities due to cross-lingual knowledge transfer and biases embedded in both pretraining and fine-tuning data. In this work, we study multilingual unlearning using the Aya-Expanse 8B model under two settings: (1) data unlearning and (2) concept unlearning. We extend benchmarks for factual knowledge and stereotypes to ten languages through translation: English, French, Arabic, Japanese, Russian, Farsi, Korean, Hindi, Hebrew, and Indonesian. These languages span five language families and a wide range of resource levels. Our experiments show that unlearning in high-resource languages is generally more stable, with asymmetric transfer effects observed between typologically related languages. Furthermore, our analysis of linguistic distances indicates that syntactic similarity is the strongest predictor of cross-lingual unlearning behavior. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_05641 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Multilingual Amnesia: On the Transferability of Unlearning in Multilingual LLMs Farashah, Alireza Dehghanpour Khandelwal, Aditi Fauchard, Marylou Shi, Zhuan Rostamzadeh, Negar Farnadi, Golnoosh Computation and Language Machine Learning As multilingual large language models become more widely used, ensuring their safety and fairness across diverse linguistic contexts presents unique challenges. While existing research on machine unlearning has primarily focused on monolingual settings, typically English, multilingual environments introduce additional complexities due to cross-lingual knowledge transfer and biases embedded in both pretraining and fine-tuning data. In this work, we study multilingual unlearning using the Aya-Expanse 8B model under two settings: (1) data unlearning and (2) concept unlearning. We extend benchmarks for factual knowledge and stereotypes to ten languages through translation: English, French, Arabic, Japanese, Russian, Farsi, Korean, Hindi, Hebrew, and Indonesian. These languages span five language families and a wide range of resource levels. Our experiments show that unlearning in high-resource languages is generally more stable, with asymmetric transfer effects observed between typologically related languages. Furthermore, our analysis of linguistic distances indicates that syntactic similarity is the strongest predictor of cross-lingual unlearning behavior. |
| title | Multilingual Amnesia: On the Transferability of Unlearning in Multilingual LLMs |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2601.05641 |