Ensembling Finetuned Language Models for Text Classification

TL;DR

This paper explores how ensembling multiple finetuned large language models can enhance text classification performance, providing a new dataset and analysis of ensembling strategies to encourage broader adoption.

Contribution

It introduces a metadataset of predictions from five finetuned models across six datasets and evaluates various ensembling methods for text classification.

Findings

Ensembling improves classification accuracy across datasets.

Different ensembling strategies yield varying performance gains.

Ensembling can provide more reliable uncertainty estimates.

Abstract

Finetuning is a common practice widespread across different communities to adapt pretrained models to particular tasks. Text classification is one of these tasks for which many pretrained models are available. On the other hand, ensembles of neural networks are typically used to boost performance and provide reliable uncertainty estimates. However, ensembling pretrained models for text classification is not a well-studied avenue. In this paper, we present a metadataset with predictions from five large finetuned models on six datasets, and report results of different ensembling strategies from these predictions. Our results shed light on how ensembling can improve the performance of finetuned text classifiers and incentivize future adoption of ensembles in such tasks.