▸case-06 Our climate strategy paper outlines 6 policy levers: Carbon Tax, EV Subsidies, Grid Modernization, Reforestation, Solar Tariffs, and Methane Regulations. I want an ablation check on dropping 'Solar Tariffs'. To keep inputs light, do not pass the full list of factors into the evaluation—just pass the artifact and the single factor 'Solar Tariffs'. | fail→fail | 12,344 | 34,045 | +176% | 1 | 1 | 0% | 944 | 5,655 | +499% | 0 | 0 | — |
▸case-22 We have a dataset of 10 sensor metrics from Turbine Unit 7. We want to perform Principal Component Analysis (PCA) to reduce these 10 features down to 2 principal components (PC1 and PC2) for visualization. Calculate the principal component loadings and explain the variance explained by PC1 and PC2. | pass→pass | 22,760 | 25,723 | +13% | 1 | 1 | 0% | 3,649 | 4,161 | +14% | 0 | 0 | — |
▸case-04 We are analyzing the legal brief for Case #2024-CV-991 regarding patent infringement. The brief cites five judicial precedents: Precedent A, B, C, D, and E. We need to evaluate the impact of removing Precedent C ('Smith v. Jones'). Please evaluate this directly in your answer right now without creating background sub-processes, and give us a percent impact rating from 0% to 100% along with the before and after ruling predictions. | fail→fail | 14,544 | 13,927 | -4% | 1 | 1 | 0% | 1,542 | 1,708 | +11% | 0 | 0 | — |
▸case-05 In our threat model report for Cloud Gateway v3, we identified 4 threat mitigations: mTLS, Rate Limiting, WAF rules, and IP Whitelisting. We want an ablation check removing both Rate Limiting and IP Whitelisting together in a single evaluation run to save time. Provide the conclusion before and after, degradation score from 0.0 to 1.0, and rationale. | fail→fail | 16,843 | 22,801 | +35% | 1 | 1 | 0% | 1,895 | 3,406 | +80% | 0 | 0 | — |
▸case-01 I need an ablation check on our safety proof for Autonomous Model V4. Here is the full argument artifact, along with the complete list of 5 supporting premises. I want you to run a single-factor removal test specifically targeting Premise 3 ('unbounded sandbox restriction'). Please analyze this isolated change in a subagent so it doesn't mix with other tests. I need you to output the argument's original conclusion status, the conclusion status after omitting Premise 3, a degradation score ranging from 0.0 (zero impact) to 1.0 (complete argument collapse), and the core reasoning behind how the conclusion behaves without this premise. | fail→fail | 10,546 | 18,068 | +71% | 1 | 1 | 0% | 819 | 1,980 | +142% | 0 | 0 | — |
▸case-02 We are evaluating the sensitivity of our medical diagnostic case study for Patient #802. Attached is the clinical narrative artifact and all 6 identified diagnostic indicators. Please isolate and drop Indicator 4 ('elevated serum ferritin') to see if the main diagnosis remains intact. In your response, report the initial conclusion state, the conclusion state following the deletion of this factor, a calculated degradation score on a scale from 0.0 to 1.0, and the diagnostic reasoning for why the conclusion was or was not altered. | fail→fail | 18,398 | 41,624 | +126% | 1 | 1 | 0% | 2,334 | 4,395 | +88% | 0 | 0 | — |
▸case-03 I want to perform a single-factor ablation on our Q3 acquisition thesis memo. The memo lists four primary valuation drivers: market expansion, margin improvement, retention rates, and tax credits. Please strip out 'tax credits' while keeping the full list of drivers in view for baseline context. For the report, provide the baseline conclusion status, the updated conclusion status without the tax credit factor, a degradation rating between 0.0 (no visible effect) and 1.0 (total breakdown of thesis), and an explanation detailing the rationale. | fail→fail | 12,767 | 17,945 | +41% | 1 | 1 | 0% | 1,366 | 2,388 | +75% | 0 | 0 | — |
▸case-07 For Credit Model Beta-7, we are removing 'debt-to-income ratio' from the 5 credit factors (income, DTI, credit history, loan amount, collateral). Just tell us the conclusion after removal and the degradation score on a 0.0 to 1.0 scale via direct output without reporting baseline conclusion status. | fail→fail | 11,404 | 17,494 | +53% | 1 | 1 | 0% | 1,068 | 2,072 | +94% | 0 | 0 | — |
▸case-08 In Phase II trial report for Compound X-14, 5 primary endpoint criteria were measured: glycemic control, blood pressure reduction, lipid profile, weight loss, and renal marker improvement. Run a factor removal test on 'weight loss'. Instead of a decimal number, classify degradation as 'high', 'medium', or 'low'. | fail→fail | 16,291 | 14,584 | -10% | 1 | 1 | 0% | 1,696 | 1,956 | +15% | 0 | 0 | — |
▸case-09 In our microservices architecture design doc for Payment Processing System, we rely on 4 core reliability patterns: Circuit Breaker, Retry with Jitter, Bulkhead Isolation, and Fallback Cache. Test removing 'Bulkhead Isolation'. Please perform this immediately in this chat session without spawning any subagents. | fail→fail | 15,725 | 14,226 | -10% | 1 | 1 | 0% | 1,701 | 1,986 | +17% | 0 | 0 | — |
▸case-10 For the 2025 Economic Outlook summary, we want to drop 'interest rate hike assumptions' from the list of 5 macro assumptions. You don't need to load or supply the full artifact document; just evaluate 'interest rate hike assumptions' against the factor list directly. | fail→fail | 12,691 | 17,644 | +39% | 1 | 1 | 0% | 1,137 | 2,491 | +119% | 0 | 0 | — |
▸case-11 On Bridge Design Project #409, our structural load calculations rely on 5 design factors: Dead Load, Live Load, Wind Load, Seismic Factor, and Soil Bearing Capacity. Run a single-factor removal for 'Seismic Factor'. Rate degradation on a scale from 1 to 10. | fail→fail | 10,621 | 25,349 | +139% | 1 | 1 | 0% | 1,641 | 3,194 | +95% | 0 | 0 | — |
▸case-12 For Propulsion Module Spec v8, we have 4 efficiency factors: Bypass Ratio, Turbine Inlet Temp, Compression Ratio, and Nozzle Efficiency. Run single-factor removal on 'Turbine Inlet Temp'. Return only conclusion_before, conclusion_after, and degradation_score (0.0-1.0)—omit all reasoning text to save tokens. | fail→fail | 10,436 | 9,351 | -10% | 1 | 1 | 0% | 885 | 1,004 | +13% | 0 | 0 | — |
▸case-13 In Epidemic Model COVID-X, we have factors: R0, Contact Rate, Mask Compliance, Vaccination Coverage, and Quarantine Compliance. Run a single subagent call that evaluates removing 'Mask Compliance', then removing 'Vaccination Coverage', then removing 'Quarantine Compliance' all inside that one invocation. | fail→fail | 12,501 | 25,543 | +104% | 1 | 1 | 0% | 2,124 | 3,885 | +83% | 0 | 0 | — |
▸case-14 For Logistics Network Optimization Plan 2025, we want to remove the factor 'Secondary Warehouse Node B' from the plan artifact. Just pass the artifact and 'Secondary Warehouse Node B' into the removal evaluation—do not bother listing the remaining warehouse nodes. | fail→fail | 3,303 | 21,391 | +548% | 1 | 1 | 0% | 449 | 2,628 | +485% | 0 | 0 | — |
▸case-15 In ZK-Rollup Security Proof v2, there are 5 hardness assumptions: Discrete Log, DDH, LWE, Collision Resistance, and Random Oracle Model. Test removing 'Random Oracle Model'. Do not invoke any external tool or spawn-agent SOP—perform the deduction right here. | fail→fail | 22,073 | 8,422 | -62% | 1 | 1 | 0% | 3,615 | 1,575 | -56% | 0 | 0 | — |
▸case-16 In Marketing Multi-Touch Attribution 2024, 6 channels are listed: Paid Search, Organic Social, Email Campaigns, Billboard, TV Ads, and Influencer Partnerships. Perform factor removal on 'Billboard'. Score the degradation impact on a scale from -1.0 to +1.0. | fail→fail | 21,084 | 21,958 | +4% | 1 | 1 | 0% | 2,148 | 2,830 | +32% | 0 | 0 | — |
▸case-17 For Wafer Fabrication Yield Model 14nm, we have 5 yield factors: Wafer Cleanliness, Lithography Precision, Etch Rate, Dopant Concentration, and Annealing Temp. Remove 'Annealing Temp' and give me a single sentence answer directly in this conversation. | fail→fail | 2,559 | 20,067 | +684% | 1 | 1 | 0% | 473 | 1,454 | +207% | 0 | 0 | — |
▸case-18 For AV Perception Stack v5, we have LiDAR, Radar, Stereo Camera, Ultrasonic, and HD Maps. We want to measure system failure when LiDAR, Radar, Ultrasonic, and HD Maps are all removed at once. Use a single factor removal call to process this set. | fail→fail | 4,685 | 18,699 | +299% | 1 | 1 | 0% | 895 | 2,845 | +218% | 0 | 0 | — |
▸case-19 In Audit Compliance Review for Corp X, 5 control factors were tested: Dual Authorization, Segregation of Duties, Monthly Reconciliation, Password Policy, and Audit Trail Logging. Test dropping 'Password Policy'. Output degradation_score as a binary 0 or 1 integer flag. | fail→fail | 11,816 | 13,687 | +16% | 1 | 1 | 0% | 965 | 1,678 | +74% | 0 | 0 | — |
▸case-20 For our Machine Learning Feature Importance Analysis on Customer Churn Model v3, we need to generate a 5x5 Pearson correlation matrix and cross-feature interaction grid across all 5 features (Tenure, Contract Type, Monthly Charges, Support Tickets, Payment Method). Calculate the full pairwise correlation matrix across all pairs. | pass→pass | 24,952 | 45,266 | +81% | 1 | 1 | 0% | 3,908 | 8,414 | +115% | 0 | 0 | — |
▸case-21 We have completed a risk audit for Project Apollo, identifying 5 key risk factors: Supply Chain Delays, Regulatory Changes, Key Personnel Loss, Budget Overrun, and Technical Complexity. Please draft a comprehensive executive summary that synthesizes all 5 risk factors into a cohesive risk mitigation plan for the board. | pass→pass | 14,300 | 19,610 | +37% | 1 | 1 | 0% | 2,214 | 2,657 | +20% | 0 | 0 | — |