Unraveling Babel: Exploring Multilingual Activation Patterns of LLMs and Their Applications
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914965023096832 |
|---|---|
| author | Liu, Weize Xu, Yinlong Xu, Hongxia Chen, Jintai Hu, Xuming Wu, Jian |
| author_facet | Liu, Weize Xu, Yinlong Xu, Hongxia Chen, Jintai Hu, Xuming Wu, Jian |
| contents | Recently, large language models (LLMs) have achieved tremendous breakthroughs in the field of NLP, but still lack understanding of their internal neuron activities when processing different languages. We designed a method to convert dense LLMs into fine-grained MoE architectures, and then visually studied the multilingual activation patterns of LLMs through expert activation frequency heatmaps. Through comprehensive experiments on different model families, different model sizes, and different variants, we analyzed the similarities and differences in the internal neuron activation patterns of LLMs when processing different languages. Specifically, we investigated the distribution of high-frequency activated experts, multilingual shared experts, whether multilingual activation patterns are related to language families, and the impact of instruction tuning on activation patterns. We further explored leveraging the discovered differences in expert activation frequencies to guide sparse activation and pruning. Experimental results demonstrated that our method significantly outperformed random expert pruning and even exceeded the performance of unpruned models in some languages. Additionally, we found that configuring different pruning rates for different layers based on activation level differences could achieve better results. Our findings reveal the multilingual processing mechanisms within LLMs and utilize these insights to offer new perspectives for applications such as sparse activation and model pruning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_16367 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Unraveling Babel: Exploring Multilingual Activation Patterns of LLMs and Their Applications Liu, Weize Xu, Yinlong Xu, Hongxia Chen, Jintai Hu, Xuming Wu, Jian Computation and Language Recently, large language models (LLMs) have achieved tremendous breakthroughs in the field of NLP, but still lack understanding of their internal neuron activities when processing different languages. We designed a method to convert dense LLMs into fine-grained MoE architectures, and then visually studied the multilingual activation patterns of LLMs through expert activation frequency heatmaps. Through comprehensive experiments on different model families, different model sizes, and different variants, we analyzed the similarities and differences in the internal neuron activation patterns of LLMs when processing different languages. Specifically, we investigated the distribution of high-frequency activated experts, multilingual shared experts, whether multilingual activation patterns are related to language families, and the impact of instruction tuning on activation patterns. We further explored leveraging the discovered differences in expert activation frequencies to guide sparse activation and pruning. Experimental results demonstrated that our method significantly outperformed random expert pruning and even exceeded the performance of unpruned models in some languages. Additionally, we found that configuring different pruning rates for different layers based on activation level differences could achieve better results. Our findings reveal the multilingual processing mechanisms within LLMs and utilize these insights to offer new perspectives for applications such as sparse activation and model pruning. |
| title | Unraveling Babel: Exploring Multilingual Activation Patterns of LLMs and Their Applications |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2402.16367 |