On the Limitations of Language Targeted Pruning: Investigating the Calibration Language Impact in Multilingual LLM Pruning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kurz, Simon, Chen, Jian-Jia, Flek, Lucie, Zhao, Zhixue
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912713819553792
author Kurz, Simon
Chen, Jian-Jia
Flek, Lucie
Zhao, Zhixue
author_facet Kurz, Simon
Chen, Jian-Jia
Flek, Lucie
Zhao, Zhixue
contents Recent advances in large language model (LLM) pruning have shown state-of-the-art (SotA) compression results in post-training and retraining-free settings while maintaining high predictive performance. However, previous research mainly considered calibrating based on English text, despite the multilingual nature of modern LLMs and their frequent use in non-English languages. This analysis paper conducts an in-depth investigation of the performance and internal representation changes associated with pruning multilingual language models for monolingual applications. We present the first comprehensive empirical study, comparing different calibration languages for pruning multilingual models across diverse languages, tasks, models, and SotA pruning techniques. We further analyze the latent subspaces, pruning masks, and individual neurons within pruned models. Our results reveal that while calibration on the target language effectively retains perplexity and yields high signal-to-noise ratios, it does not consistently improve downstream task performance. Further analysis of internal representations at three different levels highlights broader limitations of current pruning approaches: While they effectively preserve dominant information like language-specific features, this is insufficient to counteract the loss of nuanced, language-agnostic features that are crucial for knowledge retention and reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2408_14398
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle On the Limitations of Language Targeted Pruning: Investigating the Calibration Language Impact in Multilingual LLM Pruning
Kurz, Simon
Chen, Jian-Jia
Flek, Lucie
Zhao, Zhixue
Computation and Language
Artificial Intelligence
Machine Learning
Recent advances in large language model (LLM) pruning have shown state-of-the-art (SotA) compression results in post-training and retraining-free settings while maintaining high predictive performance. However, previous research mainly considered calibrating based on English text, despite the multilingual nature of modern LLMs and their frequent use in non-English languages. This analysis paper conducts an in-depth investigation of the performance and internal representation changes associated with pruning multilingual language models for monolingual applications. We present the first comprehensive empirical study, comparing different calibration languages for pruning multilingual models across diverse languages, tasks, models, and SotA pruning techniques. We further analyze the latent subspaces, pruning masks, and individual neurons within pruned models. Our results reveal that while calibration on the target language effectively retains perplexity and yields high signal-to-noise ratios, it does not consistently improve downstream task performance. Further analysis of internal representations at three different levels highlights broader limitations of current pruning approaches: While they effectively preserve dominant information like language-specific features, this is insufficient to counteract the loss of nuanced, language-agnostic features that are crucial for knowledge retention and reasoning.
title On the Limitations of Language Targeted Pruning: Investigating the Calibration Language Impact in Multilingual LLM Pruning
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2408.14398