AutoDojo: Adaptive Attacks Expose Superficial Defenses and User-Underspecification Limits in LLM Agents 文章

ArXiv CS.AI2026-06-16NEWSen作者: Xinhang Ma, Taoran Li, Chaowei Xiao, Zhiyuan Yu, Ning Zhang, Yevgeniy Vorobeychik
查看原文 →
关系图谱

AutoDojo: Adaptive Attacks Expose Superficial Defenses and User-Underspecification Limits in LLM Agents · 相关技术

相关技术

LLM-powered agentsLLMIndirect Prompt Injection