Accidental Vulnerability: Factors in Fine-Tuning that Shift Model Safeguards

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Pandey, Punya Syon, Simko, Samuel, Pelrine, Kellin, Jin, Zhijing
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908847940042752
author Pandey, Punya Syon
Simko, Samuel
Pelrine, Kellin
Jin, Zhijing
author_facet Pandey, Punya Syon
Simko, Samuel
Pelrine, Kellin
Jin, Zhijing
contents As large language models (LLMs) gain popularity, their vulnerability to adversarial attacks emerges as a primary concern. While fine-tuning models on domain-specific datasets is often employed to improve model performance, it can inadvertently introduce vulnerabilities within the underlying model. In this work, we investigate Accidental Vulnerability, unexpected vulnerabilities arising from characteristics of fine-tuning data. We begin by identifying potential correlation factors such as linguistic features, semantic similarity, and toxicity across multiple experimental datasets. We then evaluate the adversarial robustness of these fine-tuned models, analyzing persona shifts and interpretability traits to understand how dataset factors contribute to attack success rates. Lastly, we explore causal relationships that offer new insights into adversarial defense strategies, highlighting the crucial role of dataset design in preserving model alignment. Our code is available at https://github.com/psyonp/accidental_vulnerability.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16789
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Accidental Vulnerability: Factors in Fine-Tuning that Shift Model Safeguards
Pandey, Punya Syon
Simko, Samuel
Pelrine, Kellin
Jin, Zhijing
Computation and Language
Artificial Intelligence
Machine Learning
As large language models (LLMs) gain popularity, their vulnerability to adversarial attacks emerges as a primary concern. While fine-tuning models on domain-specific datasets is often employed to improve model performance, it can inadvertently introduce vulnerabilities within the underlying model. In this work, we investigate Accidental Vulnerability, unexpected vulnerabilities arising from characteristics of fine-tuning data. We begin by identifying potential correlation factors such as linguistic features, semantic similarity, and toxicity across multiple experimental datasets. We then evaluate the adversarial robustness of these fine-tuned models, analyzing persona shifts and interpretability traits to understand how dataset factors contribute to attack success rates. Lastly, we explore causal relationships that offer new insights into adversarial defense strategies, highlighting the crucial role of dataset design in preserving model alignment. Our code is available at https://github.com/psyonp/accidental_vulnerability.
title Accidental Vulnerability: Factors in Fine-Tuning that Shift Model Safeguards
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2505.16789