Google Cloud · Agentic Architect
Test sets and evaluation frameworks
Agentic Architect · Lesson 11 of 16 · 1:34:47
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the Google Cloud Professional Agentic Architect exam, on the official exam guide. In this lesson: why agents need evaluation, building test sets, ADK test files and evalsets with trajectory and response metrics, and Agent Platform Evals with LLM judges and custom autoraters. Includes three original practice questions.
Preview the topics
- Cymbal Air’s rebooking agent answers correctly, but sometimes charges a fee before it checks seat availability. What should its evaluation add?
- The agent may word its answer freely, as long as it means the same as the reference answer. Which ADK criterion fits?
- Cymbal Air wants every answer scored on its own brand voice: warm, brief, and never blaming the traveler, with no reference answer. What should the team use?