dw-skill-eval-verify

Verify evaluation quality by running an automated validation pass to check metrics like distinguishability, judge consistency, and calibration. This ensures scores are consistent and trustworthy before regression checks compare them against a baseline.

Baseline regression testing compares current evaluation scores against an established baseline to detect regressions. It generates a regression report to ensure model updates or workflow changes do not degrade evaluation quality.

Generate evaluation artifacts by running a guided verification pass with artifact generation enabled. This automatically produces skill-analysis.yaml, testset.yaml, scores.yaml, and regression_report.md to support audits and release processes.

Yes, you can check judge consistency and calibration automatically by running a verification pass with quality checks enabled. This validates the integrity of evaluation dimensions across different models and runs.

The best way to compare scores is by running a regression verification workflow. It automatically contrasts current scores with the latest baseline, surfacing actionable insights and generating a regression report for release readiness.

Run post-evaluation checks for regression health after running analysis and scoring to validate quality metrics. It is necessary before release processes to ensure consistent scores across models and verify no regressions occurred from recent changes.