Abstract:Automated red-teaming for LLMs often discovers narrow attack slices, missing diverse real-world threats, and yielding insufficient data for safety fine-tuning. We introduce Persona-Conditioned Adversarial Prompting (PCAP), which conditions adversarial search on diverse attacker personas (e.g., doctors, students, malicious actors) and strategy sets to explore realistic attack scenarios. By running parallel persona-conditioned searches, PCAP discovers transferable jailbreaks across different contexts and generates rich defense datasets with automatic metadata tracking. On GPT-OSS 120B, PCAP increases attack success from 57\% to 97\% while producing 2-6$\times$ more diverse prompts covering varied real-world scenarios. Critically, fine-tuning lightweight adapters on PCAP-generated data significantly improves model robustness (recall: 0.36 $\rightarrow$ 0.99, F1: 0.53 $\rightarrow$ 0.96) with minimal false positives, demonstrating a practical closed-loop approach from vulnerability discovery to automated alignment.
| Subjects: | Machine Learning (cs.LG); Cryptography and Security (cs.CR) |
| Cite as: | arXiv:2605.11730 [cs.LG] |
| (or arXiv:2605.11730v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.11730 arXiv-issued DOI via DataCite (pending registration) |
From: Cristian Morasso [view email]
[v1]
Tue, 12 May 2026 08:12:18 UTC (90 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。