Soft Label Memorization-Generalization for Natural Language Inference

TL;DR

This paper introduces SLMG, a fine-tuning method that uses soft labels to enhance the generalization of deep neural networks in natural language inference, leveraging label ambiguity as valuable information.

Contribution

The paper proposes SLMG, a novel fine-tuning approach that incorporates soft labels to improve DNN generalization without extensive additional labeling costs.

Findings

Improved NLI performance with soft label fine-tuning

Small soft label data (0.03%) boosts accuracy

Soft labels capture ambiguity, not noise

Abstract

Often when multiple labels are obtained for a training example it is assumed that there is an element of noise that must be accounted for. It has been shown that this disagreement can be considered signal instead of noise. In this work we investigate using soft labels for training data to improve generalization in machine learning models. However, using soft labels for training Deep Neural Networks (DNNs) is not practical due to the costs involved in obtaining multiple labels for large data sets. We propose soft label memorization-generalization (SLMG), a fine-tuning approach to using soft labels for training DNNs. We assume that differences in labels provided by human annotators represent ambiguity about the true label instead of noise. Experiments with SLMG demonstrate improved generalization performance on the Natural Language Inference (NLI) task. Our experiments show that by injecting a small percentage of soft label training data (0.03% of training set size) we can improve generalization performance over several baselines.