Just as Humans Need Vaccines, So Do Models: Model Immunization to Combat Falsehoods

TL;DR

This paper introduces model immunization, a training method that uses curated falsehood-correction pairs to reduce misinformation in large language models, improving truthfulness without sacrificing overall performance.

Contribution

The paper presents a novel immunization training paradigm that directly targets falsehoods, demonstrating significant improvements in misinformation rejection and truthfulness across multiple models.

Findings

Immunization improves TruthfulQA accuracy by 12 points.

Misinformation rejection rates increase by 30 points.

The approach maintains overall model capability.

Abstract

Large language models (LLMs) reproduce misinformation not by memorizing false facts alone, but by learning the linguistic patterns that make falsehoods persuasive, such as hedging, false presuppositions, and fabricated citations. We propose model immunization, a training paradigm based on supervised fine-tuning over curated (false claim, correction) pairs, injected as small vaccine doses (5 to 10% of tokens) alongside truthful data. Unlike post-hoc filtering or preference-based alignment, immunization introduces direct negative supervision on labeled falsehoods. Across four open weight model families, this approach improves TruthfulQA accuracy by 12 points and increases misinformation rejection rates by 30 points, while preserving overall model capability. We further outline key design requirements, including dosage, labeling, quarantine, and diversity and advocate for standardized vaccine corpora and benchmarks to evaluate generalization. These findings position immunization as a practical and scalable component of responsible LLM development.