Comparison between parameter-efficient techniques and full fine-tuning: A case study on multilingual news article classification
This work addresses the efficiency and effectiveness of fine-tuning for complex multilingual classification tasks, providing practical insights for researchers and practitioners, though it is incremental as it builds on existing parameter-efficient methods.
The study compared parameter-efficient fine-tuning techniques (Adapters and LoRA) to full fine-tuning on multilingual news article classification tasks, finding that these methods can maintain or improve performance while reducing computational costs, with specific gains observed across different languages and data scenarios.
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 classes and classification difficulty), some of which have limited training data. In addition, we conduct in-depth analyses of their efficacy across different training scenarios (training on the original multilingual data; on the translations into English; and on a subset of English-only data) and different languages. Our findings provide valuable insights into the applicability of the parameter-efficient fine-tuning techniques, particularly to complex multilingual and multilabel classification tasks.