First Train to Generate, then Generate to Train: UnitedSynT5 for Few-Shot NLI

TL;DR

This paper introduces UnitedSynT5, a synthetic data augmentation method for NLI that improves accuracy on multiple datasets by generating and integrating additional premise-hypothesis pairs, surpassing previous state-of-the-art results.

Contribution

The paper presents UnitedSynT5, a novel synthetic data augmentation approach using a T5-based generator to enhance NLI training data and improve model performance.

Findings

Achieved 94.7% accuracy on SNLI dataset.

Surpassed previous SOTA models in NLI accuracy.

Demonstrated effectiveness of synthetic data augmentation in NLI.

Abstract

Natural Language Inference (NLI) tasks require identifying the relationship between sentence pairs, typically classified as entailment, contradiction, or neutrality. While the current state-of-the-art (SOTA) model, Entailment Few-Shot Learning (EFL), achieves a 93.1% accuracy on the Stanford Natural Language Inference (SNLI) dataset, further advancements are constrained by the dataset's limitations. To address this, we propose a novel approach leveraging synthetic data augmentation to enhance dataset diversity and complexity. We present UnitedSynT5, an advanced extension of EFL that leverages a T5-based generator to synthesize additional premise-hypothesis pairs, which are rigorously cleaned and integrated into the training data. These augmented examples are processed within the EFL framework, embedding labels directly into hypotheses for consistency. We train a GTR-T5-XL model on this expanded dataset, achieving a new benchmark of 94.7% accuracy on the SNLI dataset, 94.0% accuracy on the E-SNLI dataset, and 92.6% accuracy on the MultiNLI dataset, surpassing the previous SOTA models. This research demonstrates the potential of synthetic data augmentation in improving NLI models, offering a path forward for further advancements in natural language understanding tasks.