Just as Humans Need Vaccines, So Do Models: Model Immunization to Combat Falsehoods
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866917376283377664 |
|---|---|
| author | Raza, Shaina Qureshi, Rizwan Farooq, Azib Lotif, Marcelo Chadha, Aman Pandya, Deval Emmanouilidis, Christos |
| author_facet | Raza, Shaina Qureshi, Rizwan Farooq, Azib Lotif, Marcelo Chadha, Aman Pandya, Deval Emmanouilidis, Christos |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_17870 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Just as Humans Need Vaccines, So Do Models: Model Immunization to Combat Falsehoods Raza, Shaina Qureshi, Rizwan Farooq, Azib Lotif, Marcelo Chadha, Aman Pandya, Deval Emmanouilidis, Christos Computation and Language 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. |
| title | Just as Humans Need Vaccines, So Do Models: Model Immunization to Combat Falsehoods |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2505.17870 |