Applying SoftTriple Loss for Supervised Language Model Fine Tuning

TL;DR

This paper proposes a new loss function, TripleEntropy, to enhance supervised fine-tuning of language models, showing consistent improvements especially on smaller datasets.

Contribution

Introduction of the TripleEntropy loss function that improves classification performance during language model fine-tuning over existing methods.

Findings

Steady performance gains across multiple datasets

Greater improvements observed with smaller training datasets

Enhanced robustness of fine-tuned models

Abstract

We introduce a new loss function TripleEntropy, to improve classification performance for fine-tuning general knowledge pre-trained language models based on cross-entropy and SoftTriple loss. This loss function can improve the robust RoBERTa baseline model fine-tuned with cross-entropy loss by about (0.02% - 2.29%). Thorough tests on popular datasets indicate a steady gain. The fewer samples in the training dataset, the higher gain -- thus, for small-sized dataset it is 0.78%, for medium-sized -- 0.86% for large -- 0.20% and for extra-large 0.04%.