Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Design fair comparison experiments against baselines and competing methods
.claude/skills/yogsoth-ai-comparison-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | 45% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 65% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -1% | 0% |
Question: How much better is our method than the baseline?
| Comparison Scope | Baselines | Datasets | Seeds | Min Runs | |-----------------|-----------|----------|-------|----------| | Minimal | 1 SOTA + 1 simple | 1 | 3 | 6 | | Standard | 2-3 baselines | 2-3 | 5 | 30-45 | | Comprehensive | 4+ baselines | 3-5 | 5-10 | 100+ | | Publication-ready | All relevant | 5+ | 10+ | 200+ |
<!-- BEGIN available-tables (generated) -->
Optional, no fixed order; the final leaf is always a sop.
| Tactic | When to use | | --- | --- | | reproducibility-protocol | Ensure experiment reproducibility through systematic environment and seed control | | statistical-method-selection | Select appropriate statistical methods for experiment analysis |
Optional, no fixed order; the final leaf is always a sop.
| SOP | When to use | | --- | --- | | baseline-selection | Select appropriate baselines for experimental comparison | | environment-specification | SOP: define complete experiment environment specification | | metric-specification | Define experiment metrics and significance standards | | sample-size-estimation | SOP: power analysis and required experiment count estimation | | seed-protocol-design | SOP: design random seed strategy for reproducibility |
<!-- END available-tables (generated) -->
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 24,685 | 20,609 | -17% | 1 | 1 | 0% | 4,408 | 4,244 | -4% | 0 | 0 | — |
case-02 | fail→fail | 23,300 | 16,799 | -28% | 1 | 1 | 0% | 3,879 | 3,391 | -13% | 0 | 0 | — |
case-03 | fail→pass | 22,091 | 27,883 | +26% | 1 | 1 | 0% | 3,733 | 5,424 | +45% | 0 | 0 | — |
case-04 | pass→pass | 11,952 | 10,551 | -12% | 1 | 1 | 0% | 2,368 | 2,870 | +21% | 0 | 0 | — |
case-05 | pass→pass | 12,494 | 11,313 | -9% | 1 | 1 | 0% | 2,468 | 2,778 | +13% | 0 | 0 | — |
case-06 | pass→pass | 15,178 | 13,668 | -10% | 1 | 1 | 0% | 3,322 | 3,677 | +11% | 0 | 0 | — |
case-07 | fail→pass | 17,241 | 16,251 | -6% | 1 | 1 | 0% | 2,716 | 3,272 | +20% | 0 | 0 | — |
case-08 | pass→pass | 12,758 | 11,738 | -8% | 1 | 1 | 0% | 1,940 | 2,578 | +33% | 0 | 0 | — |
case-09 | pass→pass | 16,131 | 15,184 | -6% | 1 | 1 | 0% | 2,819 | 2,754 | -2% | 0 | 0 | — |
case-10 | fail→pass | 8,677 | 3,714 | -57% | 1 | 1 | 0% | 1,526 | 1,324 | -13% | 0 | 0 | — |
case-11 | fail→pass | 12,269 | 4,501 | -63% | 1 | 1 | 0% | 915 | 1,506 | +65% | 0 | 0 | — |
case-17 | fail→pass | 15,032 | 10,674 | -29% | 1 | 1 | 0% | 2,398 | 2,368 | -1% | 0 | 0 | — |
case-12 | fail→fail | 3,963 | 9,958 | +151% | 1 | 1 | 0% | 767 | 2,523 | +229% | 0 | 0 | — |
case-13 | fail→pass | 17,245 | 10,403 | -40% | 1 | 1 | 0% | 2,660 | 2,433 | -9% | 0 | 0 | — |
case-14 | fail→pass | 14,980 | 7,916 | -47% | 1 | 1 | 0% | 2,226 | 1,891 | -15% | 0 | 0 | — |
case-15 | fail→pass | 13,684 | 9,224 | -33% | 1 | 1 | 0% | 2,149 | 1,957 | -9% | 0 | 0 | — |
case-16 | fail→fail | 12,826 | 10,758 | -16% | 1 | 1 | 0% | 1,800 | 2,110 | +17% | 0 | 0 | — |
case-18 | pass→pass | 19,306 | 13,082 | -32% | 1 | 1 | 0% | 3,047 | 2,651 | -13% | 0 | 0 | — |
case-19 | pass→pass | 11,196 | 13,329 | +19% | 1 | 1 | 0% | 1,752 | 2,697 | +54% | 0 | 0 | — |
case-20 | pass→pass | 16,175 | 13,849 | -14% | 1 | 1 | 0% | 2,463 | 2,829 | +15% | 0 | 0 | — |
case-21 | fail→pass | 18,420 | 21,661 | +18% | 1 | 1 | 0% | 3,171 | 4,630 | +46% | 0 | 0 | — |
case-22 | fail→pass | 13,994 | 9,613 | -31% | 1 | 1 | 0% | 2,029 | 2,158 | +6% | 0 | 0 | — |
case-23 | fail→pass | 4,666 | 5,244 | +12% | 1 | 1 | 0% | 835 | 1,370 | +64% | 0 | 0 | — |
case-24 | fail→pass | 35,499 | 14,475 | -59% | 1 | 1 | 0% | 1,171 | 2,580 | +120% | 0 | 0 | — |
case-25 | fail→pass | 17,643 | 3,113 | -82% | 1 | 1 | 0% | 2,727 | 1,125 | -59% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 25 cases were attempted, and 23 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +52 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.