[2508.18088] How Quantization Shapes Bias in Large Language Models

Computer Science > Computation and Language arXiv:2508.18088 (cs) [Submitted on 25 Aug 2025 (v1), last revised 5 Mar 2026 (this version, v3)] Title:How Quantization Shapes Bias in Large Language Models Authors:Federico Marcuzzi, Xuefei Ning, Roy Schwartz, Iryna Gurevych View a PDF of the paper titled How Quantization Shapes Bias in Large Language Models, by Federico Marcuzzi and 3 other authors View PDF Abstract:This work presents a comprehensive evaluation of how quantization affects model bias, with particular attention to its impact on individual demographic subgroups. We focus on weight and activation quantization strategies and examine their effects across a broad range of bias types, including stereotypes, fairness, toxicity, and sentiment. We employ both probability- and generated text-based metrics across 13 benchmarks and evaluate models that differ in architecture family and reasoning ability. Our findings show that quantization has a nuanced impact on bias: while it can reduce model toxicity and does not significantly impact sentiment, it tends to slightly increase stereotypes and unfairness in generative tasks, especially under aggressive compression. These trends are generally consistent across demographic categories and subgroups, and model types, although their magnitude depends on the specific setting. Overall, our results highlight the importance of carefully balancing efficiency and ethical considerations when applying quantization in practice. Subjects: ...