Beyond English-Centric LLMs: What Language Do Multilingual Language Models Think in?
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911996334571520 |
|---|---|
| author | Zhong, Chengzhi Cheng, Fei Liu, Qianying Jiang, Junfeng Wan, Zhen Chu, Chenhui Murawaki, Yugo Kurohashi, Sadao |
| author_facet | Zhong, Chengzhi Cheng, Fei Liu, Qianying Jiang, Junfeng Wan, Zhen Chu, Chenhui Murawaki, Yugo Kurohashi, Sadao |
| contents | In this study, we investigate whether non-English-centric LLMs, despite their strong performance, `think' in their respective dominant language: more precisely, `think' refers to how the representations of intermediate layers, when un-embedded into the vocabulary space, exhibit higher probabilities for certain dominant languages during generation. We term such languages as internal $\textbf{latent languages}$.
We examine the latent language of three typical categories of models for Japanese processing: Llama2, an English-centric model; Swallow, an English-centric model with continued pre-training in Japanese; and LLM-jp, a model pre-trained on balanced English and Japanese corpora. Our empirical findings reveal that, unlike Llama2 which relies exclusively on English as the internal latent language, Japanese-specific Swallow and LLM-jp employ both Japanese and English, exhibiting dual internal latent languages. For any given target language, the model preferentially activates the latent language most closely related to it. In addition, we explore how intermediate layers respond to questions involving cultural conflicts between latent internal and target output languages. We further explore how the language identity shifts across layers while keeping consistent semantic meaning reflected in the intermediate layer representations.
This study deepens the understanding of non-English-centric large language models, highlighting the intricate dynamics of language representation within their intermediate layers. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_10811 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Beyond English-Centric LLMs: What Language Do Multilingual Language Models Think in? Zhong, Chengzhi Cheng, Fei Liu, Qianying Jiang, Junfeng Wan, Zhen Chu, Chenhui Murawaki, Yugo Kurohashi, Sadao Computation and Language Artificial Intelligence In this study, we investigate whether non-English-centric LLMs, despite their strong performance, `think' in their respective dominant language: more precisely, `think' refers to how the representations of intermediate layers, when un-embedded into the vocabulary space, exhibit higher probabilities for certain dominant languages during generation. We term such languages as internal $\textbf{latent languages}$. We examine the latent language of three typical categories of models for Japanese processing: Llama2, an English-centric model; Swallow, an English-centric model with continued pre-training in Japanese; and LLM-jp, a model pre-trained on balanced English and Japanese corpora. Our empirical findings reveal that, unlike Llama2 which relies exclusively on English as the internal latent language, Japanese-specific Swallow and LLM-jp employ both Japanese and English, exhibiting dual internal latent languages. For any given target language, the model preferentially activates the latent language most closely related to it. In addition, we explore how intermediate layers respond to questions involving cultural conflicts between latent internal and target output languages. We further explore how the language identity shifts across layers while keeping consistent semantic meaning reflected in the intermediate layer representations. This study deepens the understanding of non-English-centric large language models, highlighting the intricate dynamics of language representation within their intermediate layers. |
| title | Beyond English-Centric LLMs: What Language Do Multilingual Language Models Think in? |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2408.10811 |