Muppet: Massive Multi-task Representations with Pre-Finetuning

TL;DR

This paper introduces pre-finetuning, a large-scale multi-task learning stage that enhances language model representations, leading to better generalization, improved performance across diverse tasks, and increased sample efficiency during fine-tuning.

Contribution

It presents a novel pre-finetuning approach involving extensive multi-task learning to improve language model generalization and performance.

Findings

Pre-finetuning improves performance on various NLP tasks.

Large-scale multi-tasking is essential for effective pre-finetuning.

Pre-finetuning enhances sample efficiency during fine-tuning.

Abstract

We propose pre-finetuning, an additional large-scale learning stage between language model pre-training and fine-tuning. Pre-finetuning is massively multi-task learning (around 50 datasets, over 4.8 million total labeled examples), and is designed to encourage learning of representations that generalize better to many different tasks. We show that pre-finetuning consistently improves performance for pretrained discriminators (e.g.~RoBERTa) and generation models (e.g.~BART) on a wide range of tasks (sentence prediction, commonsense reasoning, MRC, etc.), while also significantly improving sample efficiency during fine-tuning. We also show that large-scale multi-tasking is crucial; pre-finetuning can hurt performance when few tasks are used up until a critical point (usually above 15) after which performance improves linearly in the number of tasks.