Muppet: Massive Multi-task Representations with Pre-Finetuning
This addresses the challenge of enhancing task-specific adaptation for NLP practitioners, though it is incremental as it builds on existing pre-training and fine-tuning paradigms.
The paper tackles the problem of improving generalization and sample efficiency in language models by introducing pre-finetuning, a large-scale multi-task learning stage between pre-training and fine-tuning, which consistently boosts performance across various tasks and models, with critical improvements observed when using over 15 tasks.
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.