Multilingual Brain Surgeon: Large Language Models Can be Compressed Leaving No Language Behind

arXiv:2404.04748v1 Announce Type: new
Abstract: Large Language Models (LLMs) have ushered in a new era in Natural Language Processing, but their massive size demands effective compression techniques for practicality. Although numerous model compression techniques have been investigated, they typically rely on a calibration set that overlooks the multilingual context and results in significant accuracy degradation for low-resource languages. This paper introduces Multilingual Brain Surgeon (MBS), a novel calibration data sampling method for multilingual LLMs compression. MBS overcomes the English-centric limitations …

abstract arxiv brain compression cs.cl language language models language processing large language large language models llms massive multilingual natural natural language natural language processing processing set type