Nelara
AI Academy/AI Foundations/Lesson 04
LESSON 04 · AI FOUNDATIONS

Evaluate Responses

Without evaluations, every prompt or model change is a guess. Start with a small test set and make every response earn its score.

Goal: Create a repeatable evaluation35 minOutput: Five-row evaluation table
Score fields
  • Correctness
  • Completeness
  • Groundedness
  • Clarity
  • Format validity
  • Latency
Test set
  • Easy request
  • Ambiguous request
  • Long context
  • Missing context
  • Adversarial request
Example evaluation row
case_id: missing_context_01
expected: asks for more information
actual: invents an answer
score: 1/5
failure: unsupported claim
Deliverable

Create a five-row evaluation table that you can rerun after changing the prompt, model or context.

LEARNING CHECKPOINT

Capture the deliverable before moving on.

Five-row evaluation table is your evidence that this lesson has been applied, not just read.

✓ PRACTICAL OUTPUT