quiz · 4 min

Evaluation & Testing Check

  1. Why are task-success metrics alone not sufficient for agentic systems?
  2. What is eval overfitting, and what is the prescribed fix?
  3. In the eval harness built in lesson 31, what is the core loop structure?
  4. According to lesson 32, what is the correct way to handle non-deterministic agent outputs in evaluation?
  5. What is eval-set drift, and why does a stable golden-set pass rate not protect against it?