{"slug":"majiayu000-ai-eval-design-and-iteration","source_name":"majiayu000/ai-eval-design-and-iteration","name":"Majiayu000/AI Eval Design And Iteration","description":"Develop \"quizzes\" (evals) to measure model performance on specific tasks. Use these benchmarks to guide fine-tuning, determine product UX patterns, and track performance improvements over time. Use this when launching a new AI feature, switching between model versions, or optimizing for high-stakes accuracy.","version":1,"lift":{"pass_rate_delta_pts":30.43,"pass_rate_pct":69.6,"total_cases":23,"passed_cases":16,"tokens_delta_pct":13.2,"turns_delta_pct":0,"verdict":"mixed","benchmark_model":"gemini-3.6-flash","grading_method":"judged","completed_at":"2026-09-03T21:00:27.042461+00:00"},"skill_score":0.6957,"benchmark_models":[{"model":"gemini-3.6-flash","headline":true,"delta_pts":30.43,"with_pass_pct":69.6,"without_pass_pct":39.1,"tokens_delta_pct":13.2,"turns_delta_pct":0,"total_cases":23,"cases_aggregated":23,"verdict":"mixed","never_hurt":true,"completed_at":"2026-09-03T21:00:27.042461+00:00","run_id":"dca11329-b2fb-4b88-ad69-bf7ed77f4ca1","latest_attempt":{"id":"dca11329-b2fb-4b88-ad69-bf7ed77f4ca1","runner_model":"gemini-3.6-flash","overall_verdict":"mixed","completed_at":"2026-09-03T21:00:27.042461+00:00","gate":null},"version_number":1,"is_latest_version":true,"gate":null}],"trust":{"skill_safety":"passed","safety_status":"clean","intent_verdict":"safe","content_status":"clean","indexable":true},"license":"MIT","install_count":0,"total_uses":0,"total_users":0,"manifest_hash":"409eff7f5a36476594fd7b8d9e6c3d72a628988ea39335c74948cde3472e4fa0","install_name":"majiayu000-ai-eval-design-and-iteration","version_id":"923895a7-8877-4e52-b4e7-e1fcfe586924","body_sha256":"409eff7f5a36476594fd7b8d9e6c3d72a628988ea39335c74948cde3472e4fa0","render_hash":"4f977ce593cd6cc069f794f32ef0ac5b711462d6718586a1644fee5256709b9c","raw_url":"https://app.decimal.ai/s/majiayu000-ai-eval-design-and-iteration/SKILL.md","scorecard_url":"https://app.decimal.ai/skills/majiayu000-ai-eval-design-and-iteration"}