Active Few-Shot Learning for Text Classification Tasks

Anonymous

Abstract: The rise of Large Language Models (LLM) in the field of natural language processing has created opportunities to utilize the power of Few-Shot Learning (FSL) methods. These methods are able to achieve acceptable performance even when working with limited training data. The goal of FSL is to effectively utilize a small number of annotated samples in the learning process. However, the performance of FSL suffers when unsuitable support samples are chosen. This problem arises due to the heavy reliance on a limited number of support samples, which hampers consistent performance improvement even with the addition of more support samples. To address this challenge, we propose an active learning-based instance selection mechanism that identifies effective support instances from the unlabeled pool and is able to work with different LLMs like BART and FLAN-T5. We have conducted several experiments on three different classification tasks. The experimental results show that our proposed method consistently improves performance for different few-shot tasks.

Paper Type: short

Research Area: Efficient/Low-Resource Methods for NLP

Contribution Types: NLP engineering experiment, Approaches to low-resource settings

Languages Studied: English