How Does Data Corruption Affect Natural Language Understanding Models? A Study on GLUE datasets

TL;DR

This study investigates how data corruption affects the performance of natural language understanding models on GLUE datasets, revealing that models often rely on cues other than genuine understanding, which questions the validity of current benchmarks.

Contribution

The paper introduces controlled data corruption transformations to evaluate whether high model performance truly reflects language understanding capabilities.

Findings

Models maintain high performance on corrupted data, indicating reliance on non-sensical cues.

Corruption transformations can assess the robustness of datasets as proper language understanding benchmarks.

Results challenge the assumption that high scores equate to strong reasoning skills.

Abstract

A central question in natural language understanding (NLU) research is whether high performance demonstrates the models' strong reasoning capabilities. We present an extensive series of controlled experiments where pre-trained language models are exposed to data that have undergone specific corruption transformations. These involve removing instances of specific word classes and often lead to non-sensical sentences. Our results show that performance remains high on most GLUE tasks when the models are fine-tuned or tested on corrupted data, suggesting that they leverage other cues for prediction even in non-sensical contexts. Our proposed data transformations can be used to assess the extent to which a specific dataset constitutes a proper testbed for evaluating models' language understanding capabilities.