Low-Rank Adaptation for Multilingual Summarization: An Empirical Study

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Whitehouse, Chenxi, Huot, Fantine, Bastings, Jasmijn, Dehghani, Mostafa, Lin, Chu-Cheng, Lapata, Mirella
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909155269279744
author Whitehouse, Chenxi
Huot, Fantine
Bastings, Jasmijn
Dehghani, Mostafa
Lin, Chu-Cheng
Lapata, Mirella
author_facet Whitehouse, Chenxi
Huot, Fantine
Bastings, Jasmijn
Dehghani, Mostafa
Lin, Chu-Cheng
Lapata, Mirella
contents Although the advancements of pre-trained Large Language Models have significantly accelerated recent progress in NLP, their ever-increasing size poses significant challenges for conventional fine-tuning, especially in memory-intensive tasks. We investigate the potential of Parameter-Efficient Fine-Tuning, focusing on Low-Rank Adaptation (LoRA), in the domain of multilingual summarization, a task that is both challenging (due to typically long inputs), and relatively unexplored. We conduct an extensive study across different data availability scenarios, including high- and low-data settings, and cross-lingual transfer, leveraging models of different sizes. Our findings reveal that LoRA is competitive with full fine-tuning when trained with high quantities of data, and excels in low-data scenarios and cross-lingual transfer. We also study different strategies for few-shot cross-lingual transfer, finding that continued LoRA tuning outperforms full fine-tuning and the dynamic composition of language-specific LoRA modules.
format Preprint
id arxiv_https___arxiv_org_abs_2311_08572
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Low-Rank Adaptation for Multilingual Summarization: An Empirical Study
Whitehouse, Chenxi
Huot, Fantine
Bastings, Jasmijn
Dehghani, Mostafa
Lin, Chu-Cheng
Lapata, Mirella
Computation and Language
Artificial Intelligence
Machine Learning
Although the advancements of pre-trained Large Language Models have significantly accelerated recent progress in NLP, their ever-increasing size poses significant challenges for conventional fine-tuning, especially in memory-intensive tasks. We investigate the potential of Parameter-Efficient Fine-Tuning, focusing on Low-Rank Adaptation (LoRA), in the domain of multilingual summarization, a task that is both challenging (due to typically long inputs), and relatively unexplored. We conduct an extensive study across different data availability scenarios, including high- and low-data settings, and cross-lingual transfer, leveraging models of different sizes. Our findings reveal that LoRA is competitive with full fine-tuning when trained with high quantities of data, and excels in low-data scenarios and cross-lingual transfer. We also study different strategies for few-shot cross-lingual transfer, finding that continued LoRA tuning outperforms full fine-tuning and the dynamic composition of language-specific LoRA modules.
title Low-Rank Adaptation for Multilingual Summarization: An Empirical Study
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2311.08572