Mitigating Label Length Bias in Large Language Models
Sanz-Guerrero, Mario, von der Wense, Katharina
–arXiv.org Artificial Intelligence
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.
arXiv.org Artificial Intelligence
Nov-19-2025
- Country:
- Asia
- China > Hong Kong (0.04)
- Middle East > UAE
- Abu Dhabi Emirate > Abu Dhabi (0.04)
- Singapore (0.04)
- Europe
- Belgium > Brussels-Capital Region
- Brussels (0.04)
- Germany > Rheinland-Pfalz
- Mainz (0.04)
- Ireland > Leinster
- County Dublin > Dublin (0.04)
- Belgium > Brussels-Capital Region
- North America
- Canada > Ontario
- Toronto (0.04)
- United States
- California > San Francisco County
- San Francisco (0.14)
- Colorado > Boulder County
- Boulder (0.04)
- Minnesota > Hennepin County
- Minneapolis (0.14)
- New Mexico > Bernalillo County
- Albuquerque (0.04)
- Washington > King County
- Seattle (0.04)
- California > San Francisco County
- Canada > Ontario
- Asia
- Genre:
- Research Report > New Finding (0.93)
- Industry:
- Education (0.48)
- Information Technology (0.46)
- Technology: