Discovering the Hidden Role of Gini Index In Prompt-based Classification
This work addresses the critical issue of class imbalance in classification tasks, which is particularly important for applications where minority classes are most important, though it appears incremental in its approach.
The paper tackles the problem of accuracy disparities between majority and minority classes in prompt-based classification by revealing the hidden role of the Gini Index as a tool for detecting and optimizing these imbalances. It introduces a post-hoc model-agnostic bias mitigation method that significantly reduces both relative and absolute accuracy imbalances across few-shot news, biomedical, and zero-shot image classification tasks.
In classification tasks, the long-tailed minority classes usually offer the predictions that are most important. Yet these classes consistently exhibit low accuracies, whereas a few high-performing classes dominate the game. We pursue a foundational understanding of the hidden role of Gini Index as a tool for detecting and optimizing (debiasing) disparities in class accuracy, focusing on the case of prompt-based classification. We introduce the intuitions, benchmark Gini scores in real-world LLMs and vision models, and thoroughly discuss the insights of Gini not only as a measure of relative accuracy dominance but also as a direct optimization metric. Through rigorous case analyses, we first show that weak to strong relative accuracy imbalance exists in both prompt-based, text and image classification results and regardless of whether the classification is high-dimensional or low-dimensional. Then, we harness the Gini metric to propose a post-hoc model-agnostic bias mitigation method. Experimental results across few-shot news, biomedical, and zero-shot image classification show that our method significantly reduces both relative and absolute accuracy imbalances, minimizing top class relative dominance while elevating weakest classes.