Improving and Simplifying Pattern Exploiting Training
This work addresses the challenge of few-shot learning for natural language processing tasks, offering a more data-efficient method that eliminates the need for unlabeled data, though it is incremental as it builds upon existing PET techniques.
The paper tackled the problem of few-shot learning with limited labeled data by introducing ADAPET, which modifies Pattern Exploiting Training to provide denser supervision without requiring task-specific unlabeled data, resulting in improved performance on SuperGLUE benchmarks.
Recently, pre-trained language models (LMs) have achieved strong performance when fine-tuned on difficult benchmarks like SuperGLUE. However, performance can suffer when there are very few labeled examples available for fine-tuning. Pattern Exploiting Training (PET) is a recent approach that leverages patterns for few-shot learning. However, PET uses task-specific unlabeled data. In this paper, we focus on few-shot learning without any unlabeled data and introduce ADAPET, which modifies PET's objective to provide denser supervision during fine-tuning. As a result, ADAPET outperforms PET on SuperGLUE without any task-specific unlabeled data. Our code can be found at https://github.com/rrmenon10/ADAPET.