AM-Bench3 min read
AM-Bench: A theoretically-grounded benchmark for agentic misalignment
This is a brief overview of our new paper (📄) AM-Bench: A Unified Taxonomy and Evaluation Suite for Agentic Misalignment accepted at NeurIPS E&D 2026 💫OverviewWe introduce AM-Bench, a unified taxonomy and evaluation suite for agentic misalignment, a failure mode where models actively choose…
By eric-zLessWrong

Read the full story
Originally published by LessWrong on Oct 21, 2026.
Comments
Sign in to join the discussion.