Inference-Time Mitigation of Adversarial Political Bias in Large Language Models
The paper studies how adversarial prompts can push LLM-generated summaries of legislative videos toward partisan framing, and evaluates inference-time defenses such as chain-of-thought-style recursive self-correction plus preference-optimization ideas to restore political neutrality. Its main contribution is treating political bias as an adversarial prompt-injection problem and showing a large neutrality gain from a lightweight self-correction loop, though the evidence seems limited to one legislative-summary setting, Likert-style evaluation, and prompt-level mitigation rather than a broadly validated or principled robustness guarantee.
Abstract
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 is a critical step towards safer and more trustworthy Artificial Intelligence (AI). Current model alignment paradigms, such as reinforcement learning from human feedback (RLHF), make LLMs follow overarching safety instructions. However, this instruction tuning can be exploited via adversarial prompt injection and be used to generate unsafe content. In particular, political bias has not been specifically targeted by modern alignment techniques as harmful and biased content. To address this vulnerability of LLMs, we propose mitigation strategies using Chain of Thought (CoT) prompting and Direct Preference Optimization (DPO). Using a public dataset of legislative videos, we generate summaries using LLMs, inject bias via adversarial prompting and evaluate their performance on a four axis scale designed for political summarization. In this paper, we present different methods to shield LLMs against the injection of political bias. Our results demonstrate that the proposed Recursive Self-Correction approach raises model performance from a Political Neutrality Likert scale baseline of 2.14 to 4.56, averaged across all models, demonstrating effective inference-time mitigation of political bias in LLM-generated summaries.