Breaking the Language Barrier: Can Direct Inference Outperform Pre-Translation in Multilingual LLM Applications?
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911791481618432 |
|---|---|
| author | Intrator, Yotam Halfon, Matan Goldenberg, Roman Tsarfaty, Reut Eyal, Matan Rivlin, Ehud Matias, Yossi Aizenberg, Natalia |
| author_facet | Intrator, Yotam Halfon, Matan Goldenberg, Roman Tsarfaty, Reut Eyal, Matan Rivlin, Ehud Matias, Yossi Aizenberg, Natalia |
| contents | Large language models hold significant promise in multilingual applications. However, inherent biases stemming from predominantly English-centric pre-training have led to the widespread practice of pre-translation, i.e., translating non-English inputs to English before inference, leading to complexity and information loss. This study re-evaluates the need for pre-translation in the context of PaLM2 models (Anil et al., 2023), which have been established as highly performant in multilingual tasks. We offer a comprehensive investigation across 108 languages and 6 diverse benchmarks, including open-end generative tasks, which were excluded from previous similar studies. Our findings challenge the pre-translation paradigm established in prior research, highlighting the advantages of direct inference in PaLM2. Specifically, PaLM2-L consistently outperforms pre-translation in 94 out of 108 languages. These findings pave the way for more efficient and effective multilingual applications, alleviating the limitations associated with pre-translation and unlocking linguistic authenticity. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_04792 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Breaking the Language Barrier: Can Direct Inference Outperform Pre-Translation in Multilingual LLM Applications? Intrator, Yotam Halfon, Matan Goldenberg, Roman Tsarfaty, Reut Eyal, Matan Rivlin, Ehud Matias, Yossi Aizenberg, Natalia Computation and Language Machine Learning Large language models hold significant promise in multilingual applications. However, inherent biases stemming from predominantly English-centric pre-training have led to the widespread practice of pre-translation, i.e., translating non-English inputs to English before inference, leading to complexity and information loss. This study re-evaluates the need for pre-translation in the context of PaLM2 models (Anil et al., 2023), which have been established as highly performant in multilingual tasks. We offer a comprehensive investigation across 108 languages and 6 diverse benchmarks, including open-end generative tasks, which were excluded from previous similar studies. Our findings challenge the pre-translation paradigm established in prior research, highlighting the advantages of direct inference in PaLM2. Specifically, PaLM2-L consistently outperforms pre-translation in 94 out of 108 languages. These findings pave the way for more efficient and effective multilingual applications, alleviating the limitations associated with pre-translation and unlocking linguistic authenticity. |
| title | Breaking the Language Barrier: Can Direct Inference Outperform Pre-Translation in Multilingual LLM Applications? |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2403.04792 |