Mitigating Label Length Bias in Large Language Models

TL;DR

This paper introduces normalized contextual calibration (NCC), a novel method to reduce label length bias in large language models, significantly improving their accuracy and robustness in multi-token classification tasks.

Contribution

The paper proposes NCC, a new calibration technique that normalizes and calibrates predictions at the full-label level to mitigate label length bias in LLMs.

Findings

NCC improves F1 scores by up to 10% across datasets and models.

NCC reduces sensitivity to few-shot example selection.

NCC enhances confidence estimate reliability.

Abstract

Large language models (LLMs) are powerful zero- and few-shot learners. However, when predicting over a set of candidate options, LLMs suffer from label biases, and existing calibration methods overlook biases arising from multi-token class labels. We tackle an issue we call label length bias, where labels of different lengths are treated inconsistently, even after standard length normalization. To mitigate it, we propose normalized contextual calibration (NCC), an effective method that normalizes and calibrates predictions at the full-label level. NCC achieves statistically significant improvements over prior approaches across multiple datasets and models, with gains of up to 10% F1. Moreover, NCC extends bias mitigation to broader tasks such as multiple-choice question answering. Our analysis shows that, when combined with in-context learning, NCC is less sensitive to few-shot example selection, requires fewer examples for competitive performance, and produces more reliable confidence estimates. These findings highlight the importance of mitigating full-label biases to improve the performance and robustness of LLM-based methods, particularly in real-world applications where class labels naturally consist of multiple tokens.