How to Tune a Multilingual Encoder Model for Germanic Languages: A Study of PEFT, Full Fine-Tuning, and Language Adapters

TL;DR

This study examines how to tune the multilingual encoder mDeBERTa for three Germanic languages by comparing full fine-tuning, two PEFT approaches (LoRA and Pfeiffer adapters), and language adapters. Results reveal a strong dependency on language resource level and task: PEFT excels for extractive QA in German, while full fine-tuning often yields better NER performance, with Swedish and Icelandic showing mixed outcomes. Language adapters provide little to no consistent benefit, suggesting pretraining data suffices for these languages in this setup. The findings offer practical guidance for deploying multilingual DeBERTa models and point to areas for further optimization of PEFT methods and adapter configurations.

Abstract

This paper investigates the optimal use of the multilingual encoder model mDeBERTa for tasks in three Germanic languages -- German, Swedish, and Icelandic -- representing varying levels of presence and likely data quality in mDeBERTas pre-training data. We compare full fine-tuning with the parameter-efficient fine-tuning (PEFT) methods LoRA and Pfeiffer bottleneck adapters, finding that PEFT is more effective for the higher-resource language, German. However, results for Swedish and Icelandic are less consistent. We also observe differences between tasks: While PEFT tends to work better for question answering, full fine-tuning is preferable for named entity recognition. Inspired by previous research on modular approaches that combine task and language adapters, we evaluate the impact of adding PEFT modules trained on unstructured text, finding that this approach is not beneficial.