Learning to Configure Agentic AI Systems

PAC Bench: Do Foundation Models Understand Prerequisites for Executing Manipulation Policies?