AM-Bench: A Unified Taxonomy and Evaluation Suite for Agentic Misalignment↗ (원문, 새 창)

본문 미리보기

arXiv:2610.24264v1 Announce Type: new Abstract: As LLM-based agents are deployed with increasing autonomy in real-world environments, a critical failure mode emerges: agentic misalignment, where agents choose harmful actions over failure. However, existing safety benchmarks primarily evaluate whether agents comply with explicitly harmful instructions, leaving the conditions that cause agents to choose harmful actions poorly understood. To address this limitation, we propose AM-Bench, a unified