How Robust is GPT-3.5 to Predecessors? A Comprehensive Study on Language Understanding Tasks

TL;DR

This paper evaluates GPT-3.5's robustness across diverse NLP tasks and transformations, revealing significant performance drops and specific robustness challenges, which are crucial for trustworthy AI development.

Contribution

It provides a comprehensive analysis of GPT-3.5's robustness using extensive datasets and transformations, highlighting its limitations and areas for improvement.

Findings

GPT-3.5's performance drops up to 43.59% under transformations

Identifies robustness issues like instability, prompt sensitivity, and number sensitivity

Highlights the need for robustness improvements for trustworthy AI

Abstract

The GPT-3.5 models have demonstrated impressive performance in various Natural Language Processing (NLP) tasks, showcasing their strong understanding and reasoning capabilities. However, their robustness and abilities to handle various complexities of the open world have yet to be explored, which is especially crucial in assessing the stability of models and is a key aspect of trustworthy AI. In this study, we perform a comprehensive experimental analysis of GPT-3.5, exploring its robustness using 21 datasets (about 116K test samples) with 66 text transformations from TextFlint that cover 9 popular Natural Language Understanding (NLU) tasks. Our findings indicate that while GPT-3.5 outperforms existing fine-tuned models on some tasks, it still encounters significant robustness degradation, such as its average performance dropping by up to 35.74\% and 43.59\% in natural language inference and sentiment analysis tasks, respectively. We also show that GPT-3.5 faces some specific robustness challenges, including robustness instability, prompt sensitivity, and number sensitivity. These insights are valuable for understanding its limitations and guiding future research in addressing these challenges to enhance GPT-3.5's overall performance and generalization abilities.