Discovering the Hidden Role of Gini Index In Prompt-based Classification

TL;DR

This paper explores the Gini Index as a tool for detecting and reducing class accuracy disparities in prompt-based classification, demonstrating its effectiveness across various models and tasks.

Contribution

It introduces the Gini Index as a novel metric for understanding and debiasing class accuracy imbalances, along with a post-hoc bias mitigation method.

Findings

Gini scores reveal significant accuracy imbalances in prompt-based classification.

The proposed method effectively reduces class accuracy disparities.

Experimental results show improved fairness and balanced class performance.

Abstract

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.