Inference-Time Mitigation of Adversarial Political Bias in Large Language Models

Tejaswi V. Panchagnula, Bruce Coburn, Bryce J. Dietrich +3

As Large Language Models (LLMs) become the mainstay for information retrieval and summarization tasks, ensuring that they are always non-partisan and invulnerable to political bias…