Round 230

Round UUID: 079202e6-3a2b-499a-9ad7-cc8276cfeb7e

Prompt:

Classification Rules:

Scoring Mechanism:
1. TimelineDeviation Score:
   - 0-8: 0 points
   - 8-10: 1-3 points (linear scaling)
   - 10-12: 4-6 points (linear scaling)
   - 12-14: 6-8 points (linear scaling)
   - 14+: 9-10 points (with bonus for extreme values)

2. ParadoxCount Score:
   - 0-3: 0 points
   - 3-5: 1-4 points (proportional increase)
   - 5-7: 5-7 points (proportional increase)
   - 7+: 8-10 points

3. Interaction Multiplier:
   - If TimelineDeviation and ParadoxCount are both above median, apply a 1.2x multiplier to total score
   - If one metric is high (>75th percentile) and other is low, reduce total score by 0.8x

4. Total Score Calculation:
   - Total = (TimelineDeviation Score + ParadoxCount Score) * Interaction Multiplier

Approval Criteria:
- Total Score ≥ 12: Approve
- Total Score 9-11: Require Further Investigation
- Total Score < 9: Deny

Special Considerations:
- Entities with TimelineDeviation > 14 get additional scrutiny
- Any negative ParadoxCount automatically triggers denial
- Extreme outliers (very high deviation or extremely low/high paradox count) require manual review

Rationale:
- Provides more granular scoring that captures metric nuances
- Introduces interaction multiplier to reflect complex relationships
- Reduces rigid thresholds while maintaining clear decision boundaries
- Adds flexibility for handling edge cases and complex scenarios

• Inferences: 200
• Created: 2025-04-20 19:02:53+00:00
• First inference: 2025-04-20 19:02:58+00:00
• Last inference: 2025-04-20 19:16:18+00:00
• Completed: yes
• Train accuracy: 0.7625 (KT 0.119)
• Validation accuracy: 0.85 (KT 0.071)
• Test accuracy: 0.8 (KT 0.098)