Revisiting Parameter-Efficient Tuning: Are We Really There Yet?
This work critically assesses PETuning methods, revealing reliability problems that could mislead practitioners in NLP, making it an incremental but important correction.
The authors re-examined Parameter-Efficient Tuning (PETuning) methods for pretrained language models and found that under a fair evaluation protocol, PETuning does not consistently match finetuning performance in medium- and high-resource settings, with instability issues identified.
Parameter-Efficient Tuning (PETuning) methods have been deemed by many as the new paradigm for using pretrained language models (PLMs). By tuning just a fraction amount of parameters comparing to full model finetuning, PETuning methods claim to have achieved performance on par with or even better than finetuning. In this work, we take a step back and re-examine these PETuning methods by conducting the first comprehensive investigation into the training and evaluation of them. We found the problematic validation and testing practice in current studies, when accompanied by the instability nature of PETuning methods, has led to unreliable conclusions. When being compared under a truly fair evaluation protocol, PETuning cannot yield consistently competitive performance while finetuning remains to be the best-performing method in medium- and high-resource settings. We delve deeper into the cause of the instability and observed that the number of trainable parameters and training iterations are two main factors: reducing trainable parameters and prolonging training iterations may lead to higher stability in PETuning methods.