Fighting Randomness With Randomness: Mitigating Optimisation Instability of Fine-Tuning Using Ensemble and Noise Regularisation

article in a collection out of WoS and Scopus

English

While fine-tuning of pre-trained language models generally helps to overcome the lack of labelled training samples, it also displays model performance instability. This instability mainly originates from randomness in initialisation or data shuffling. To address this, researchers either modify the training process or augment the available samples, which typically results in increased computational costs. We propose a new mitigation strategy, called Delayed Ensemble with Noisy Interpolation (DENI), that leverages the strengths of ensembling, noise regularisation and model interpolation, while retaining computational efficiency. We compare DENI with 9 representative mitigation strategies across 3 models, 4 tuning strategies and 7 text classification datasets. We show that: 1) DENI outperforms the best performing mitigation strategy (Ensemble), while using only a fraction of its cost; 2) the mitigation strategies are beneficial for parameter-efficient fine-tuning (PEFT) methods, outperforming full fine-tuning in specific cases; and 3) combining DENI with data augmentation often leads to even more effective instability mitigation.

NLP in resource-constrained settings, parameter-efficient-training, data-efficient training, data augmentation, fine-tuning, mitigating randomness, ensembling

PECHER, B.; ČEGIŇ, J.; BELANEC, R.; SRBA, I.; ŠIMKO, J.; BIELIKOVÁ, M.

2. 11. 2024

Association for Computational Linguistics

Miami

979-8-8917-6168-1

Findings of the Association for Computational Linguistics: EMNLP 2024

11005

11044

40