Comparison between parameter-efficient techniques and full fine-tuning: A case study on multilingual news article classification

arXiv:2308.07282v2 Announce Type: replace
Abstract: Adapters and Low-Rank Adaptation (LoRA) are parameter-efficient fine-tuning techniques designed to make the training of language models more efficient. Previous results demonstrated that these methods can even improve performance on some classification tasks. This paper complements the existing research by investigating how these techniques influence the classification performance and computation costs compared to full fine-tuning when applied to multilingual text classification tasks (genre, framing, and persuasion techniques detection; with different input lengths, number of predicted …

abstract article arxiv case case study classification comparison cs.cl fine-tuning language language models lora low low-rank adaptation multilingual paper performance research results study tasks training type