Skip to content
masterclass Buddy
Start learning

AI Agents · Concept

Evaluation

Evaluate an agent by checking its final answer, one step, or full trajectory. Use deterministic trajectory matching for known tool paths, or an LLM judge for broader behavior.

Last updated

Syntax

evaluator({ outputs, referenceOutputs })

In-depth API docs are only for subscribers: parameters, examples, common mistakes and interview practice.

Subscribe
  • final response evaluation
  • single-step evaluation
  • trajectory evaluation
  • trajectory match
  • LLM-as-judge
  • RAG evaluation
  • summarization evaluation
  • classification and tagging evaluation

Source: Agent Evals. Checked 2026-11-03.