Cloud and provider evals

Anthropic evaluation guide

Anthropic's guide explains how to define measurable success criteria, assemble task-specific and held-out datasets, include edge cases, combine metrics, and automate grading. It is a methodology resource rather than a hosted evaluation product.

Maintainer
Anthropic
Deployment
Documentation and sample code; no managed runner
Status
Current as of September 22, 2026

What it does

Anthropic's guide explains how to define measurable success criteria, assemble task-specific and held-out datasets, include edge cases, combine metrics, and automate grading. It is a methodology resource rather than a hosted evaluation product.

  • Specific, measurable, achievable, and relevant success criteria
  • Task datasets, edge cases, held-out tests, multidimensional metrics, and automated grading
  • Examples for exact match, text similarity, ROUGE-style metrics, and LLM-based rubrics

Where it fits

Teams designing task-specific evaluation suites for Claude applications that need criteria, datasets, edge cases, grading patterns, and code examples.

Use it after defining the behavior that matters, collecting representative cases, and deciding how each case will be graded. A tool can execute and organize an eval, but it cannot decide which failures are costly for your users.

Limitations and cautions

  • This is guidance, not a runner, dashboard, dataset store, monitoring service, or SLA-backed product.
  • Teams must operate the harness, persistence, CI, monitoring, and human review, and must calibrate qualitative judges.

How to evaluate the evaluator

Before using any score as a release gate, create a small evaluator-validation set that humans have labeled. Compare false passes, false failures, disagreement by failure type, run-to-run variance, latency, and cost. Keep deterministic checks for verifiable facts, calibrate model judges against domain experts, and review raw traces when aggregate scores move.

Primary sources