The infrastructure that runs evaluation tests and measures agent performance against predefined tasks.