ContextualJailbreak: Evolutionary Red-Teaming via Simulated Conversational Priming

Abstract

Large language models (LLMs) remain vulnerable to jailbreak attacks that bypass safety alignment and elicit harmful responses. A growing body of work shows that contextual priming, where earlier turns covertly bias later replies, constitutes a powerful attack surface, with hand-crafted multi-turn scaffolds consistently outperforming single-turn manipulations on capable models. However, automated optimization-based red-teaming has remained largely limited to the single-turn setting, iterating ove

Categories

Framework mappings

Suggested from the entry's categories.

@misc{bejar2026contextualjailbreak,
  title = {{ContextualJailbreak: Evolutionary Red-Teaming via Simulated Conversational Priming}},
  author = {Mario Rodríguez Béjar and Francisco J. Cortés-Delgado and S. Braghin and Jose L. Hernández-Ramos},
  year = {2026},
  month = may,
  eprint = {2605.02647},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2605.02647}
}