Are All Prompt Components Value-Neutral? Understanding the Heterogeneous Adversarial Robustness of Dissected Prompt in Large Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912516821483520 |
|---|---|
| author | Zheng, Yujia Li, Tianhao Huang, Haotian Zeng, Tianyu Lu, Jingyu Chu, Chuangxin Huang, Yuekai Jiang, Ziyou Xiong, Qian Ge, Yuyao Li, Mingyang |
| author_facet | Zheng, Yujia Li, Tianhao Huang, Haotian Zeng, Tianyu Lu, Jingyu Chu, Chuangxin Huang, Yuekai Jiang, Ziyou Xiong, Qian Ge, Yuyao Li, Mingyang |
| contents | Prompt-based adversarial attacks have become an effective means to assess the robustness of large language models (LLMs). However, existing approaches often treat prompts as monolithic text, overlooking their structural heterogeneity-different prompt components contribute unequally to adversarial robustness. Prior works like PromptRobust assume prompts are value-neutral, but our analysis reveals that complex, domain-specific prompts with rich structures have components with differing vulnerabilities. To address this gap, we introduce PromptAnatomy, an automated framework that dissects prompts into functional components and generates diverse, interpretable adversarial examples by selectively perturbing each component using our proposed method, ComPerturb. To ensure linguistic plausibility and mitigate distribution shifts, we further incorporate a perplexity (PPL)-based filtering mechanism. As a complementary resource, we annotate four public instruction-tuning datasets using the PromptAnatomy framework, verified through human review. Extensive experiments across these datasets and five advanced LLMs demonstrate that ComPerturb achieves state-of-the-art attack success rates. Ablation studies validate the complementary benefits of prompt dissection and PPL filtering. Our results underscore the importance of prompt structure awareness and controlled perturbation for reliable adversarial robustness evaluation in LLMs. Code and data are available at https://github.com/Yujiaaaaa/PACP. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_01554 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Are All Prompt Components Value-Neutral? Understanding the Heterogeneous Adversarial Robustness of Dissected Prompt in Large Language Models Zheng, Yujia Li, Tianhao Huang, Haotian Zeng, Tianyu Lu, Jingyu Chu, Chuangxin Huang, Yuekai Jiang, Ziyou Xiong, Qian Ge, Yuyao Li, Mingyang Computation and Language Artificial Intelligence Cryptography and Security Prompt-based adversarial attacks have become an effective means to assess the robustness of large language models (LLMs). However, existing approaches often treat prompts as monolithic text, overlooking their structural heterogeneity-different prompt components contribute unequally to adversarial robustness. Prior works like PromptRobust assume prompts are value-neutral, but our analysis reveals that complex, domain-specific prompts with rich structures have components with differing vulnerabilities. To address this gap, we introduce PromptAnatomy, an automated framework that dissects prompts into functional components and generates diverse, interpretable adversarial examples by selectively perturbing each component using our proposed method, ComPerturb. To ensure linguistic plausibility and mitigate distribution shifts, we further incorporate a perplexity (PPL)-based filtering mechanism. As a complementary resource, we annotate four public instruction-tuning datasets using the PromptAnatomy framework, verified through human review. Extensive experiments across these datasets and five advanced LLMs demonstrate that ComPerturb achieves state-of-the-art attack success rates. Ablation studies validate the complementary benefits of prompt dissection and PPL filtering. Our results underscore the importance of prompt structure awareness and controlled perturbation for reliable adversarial robustness evaluation in LLMs. Code and data are available at https://github.com/Yujiaaaaa/PACP. |
| title | Are All Prompt Components Value-Neutral? Understanding the Heterogeneous Adversarial Robustness of Dissected Prompt in Large Language Models |
| topic | Computation and Language Artificial Intelligence Cryptography and Security |
| url | https://arxiv.org/abs/2508.01554 |