Improving Pre-trained Language Model Fine-tuning with Noise Stability Regularization

TL;DR

This paper introduces Layerwise Noise Stability Regularization (LNSR), a novel fine-tuning method for pre-trained language models that injects noise into hidden representations to improve generalization and reduce overfitting, validated on complex NLP tasks.

Contribution

The paper proposes a new noise-based regularization framework, LNSR, with theoretical support, outperforming existing methods on both simple and complex NLP tasks including question answering.

Findings

LNSR improves in-domain performance of language models.

LNSR enhances out-of-domain generalization.

LNSR outperforms L2-SP, Mixout, and SMART in experiments.

Abstract

The advent of large-scale pre-trained language models has contributed greatly to the recent progress in natural language processing. Many state-of-the-art language models are first trained on a large text corpus and then fine-tuned on downstream tasks. Despite its recent success and wide adoption, fine-tuning a pre-trained language model often suffers from overfitting, which leads to poor generalizability due to the extremely high complexity of the model and the limited training samples from downstream tasks. To address this problem, we propose a novel and effective fine-tuning framework, named Layerwise Noise Stability Regularization (LNSR). Specifically, we propose to inject the standard Gaussian noise or In-manifold noise and regularize hidden representations of the fine-tuned model. We first provide theoretical analyses to support the efficacy of our method. We then demonstrate the advantages of the proposed method over other state-of-the-art algorithms including L2-SP, Mixout and SMART. While these previous works only verify the effectiveness of their methods on relatively simple text classification tasks, we also verify the effectiveness of our method on question answering tasks, where the target problem is much more difficult and more training examples are available. Furthermore, extensive experimental results indicate that the proposed algorithm can not only enhance the in-domain performance of the language models but also improve the domain generalization performance on out-of-domain data.