Engineer Atlas
OverviewLearnArchitecture FinderPlaygroundFrameworksRoadmapPracticeInterview
OverviewLearnArchitecture FinderPlaygroundFrameworksRoadmapPracticeInterviewHiring GuideCheat SheetCompare
Agentic Engineering
  • Fundamentals
  • Agent Architecture
  • Tool Calling
  • RAG Engineering
  • Context Engineering
  • Memory & State
  • Planning
  • MCP — Model Context Protocol
  • Multi-Agent Systems
  • Human-in-the-Loop
  • Evaluation & Testing
  • Observability
  • Guardrails & Security
  • Reliability Engineering
  • Frameworks
Agentic/Learn/Evaluation & Testing

Evaluation & Testing

Testing probabilistic systems: golden datasets, judges, deterministic evaluators, regression.

Evaluating Agents: Testing Probabilistic Systems
▶ interactive

Evals are the test suite for systems whose output is a distribution, not a value: a fixed dataset, repeatable runs, evaluators that score traces, and a comparison against the last version.

Eval Metrics: What to Measure and How

A catalogue of agent metrics — outcome, trajectory, retrieval, efficiency and safety — with a precise definition and computation for each, so a number means the same thing across versions.

Golden Datasets

A golden dataset is the versioned set of inputs and expected outcomes every eval runs against; its coverage, hard cases and hygiene determine whether the numbers mean anything.

LLM-as-Judge

Using a model to score outputs scales evaluation to criteria code cannot express, but the judge has biases of its own and is only trustworthy after calibration against human labels.

Deterministic Evaluators

Most of what matters about an agent run can be checked by code — schemas, tool calls, trajectory shape, budgets — and those checks are exact, free and never flaky; use them before reaching for a judge.

Regression Gates and Online Evaluation

Offline evals gate changes in CI; online evaluation samples production traces, runs A/B or shadow comparisons and closes the feedback loop — and both need enough samples to separate signal from noise.

Engineer Atlas
GitHub·LinkedIn