Intrinsic Model Weaknesses: How Priming Attacks Unveil Vulnerabilities in Large Language Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866929726972493824 |
|---|---|
| author | Huang, Yuyi Zhan, Runzhe Wong, Derek F. Chao, Lidia S. Tao, Ailin |
| author_facet | Huang, Yuyi Zhan, Runzhe Wong, Derek F. Chao, Lidia S. Tao, Ailin |
| contents | Large language models (LLMs) have significantly influenced various industries but suffer from a critical flaw, the potential sensitivity of generating harmful content, which poses severe societal risks. We developed and tested novel attack strategies on popular LLMs to expose their vulnerabilities in generating inappropriate content. These strategies, inspired by psychological phenomena such as the "Priming Effect", "Safe Attention Shift", and "Cognitive Dissonance", effectively attack the models' guarding mechanisms. Our experiments achieved an attack success rate (ASR) of 100% on various open-source models, including Meta's Llama-3.2, Google's Gemma-2, Mistral's Mistral-NeMo, Falcon's Falcon-mamba, Apple's DCLM, Microsoft's Phi3, and Qwen's Qwen2.5, among others. Similarly, for closed-source models such as OpenAI's GPT-4o, Google's Gemini-1.5, and Claude-3.5, we observed an ASR of at least 95% on the AdvBench dataset, which represents the current state-of-the-art. This study underscores the urgent need to reassess the use of generative models in critical applications to mitigate potential adverse societal impacts. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_16491 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Intrinsic Model Weaknesses: How Priming Attacks Unveil Vulnerabilities in Large Language Models Huang, Yuyi Zhan, Runzhe Wong, Derek F. Chao, Lidia S. Tao, Ailin Computation and Language Large language models (LLMs) have significantly influenced various industries but suffer from a critical flaw, the potential sensitivity of generating harmful content, which poses severe societal risks. We developed and tested novel attack strategies on popular LLMs to expose their vulnerabilities in generating inappropriate content. These strategies, inspired by psychological phenomena such as the "Priming Effect", "Safe Attention Shift", and "Cognitive Dissonance", effectively attack the models' guarding mechanisms. Our experiments achieved an attack success rate (ASR) of 100% on various open-source models, including Meta's Llama-3.2, Google's Gemma-2, Mistral's Mistral-NeMo, Falcon's Falcon-mamba, Apple's DCLM, Microsoft's Phi3, and Qwen's Qwen2.5, among others. Similarly, for closed-source models such as OpenAI's GPT-4o, Google's Gemini-1.5, and Claude-3.5, we observed an ASR of at least 95% on the AdvBench dataset, which represents the current state-of-the-art. This study underscores the urgent need to reassess the use of generative models in critical applications to mitigate potential adverse societal impacts. |
| title | Intrinsic Model Weaknesses: How Priming Attacks Unveil Vulnerabilities in Large Language Models |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2502.16491 |