Confidence Policy
Defines how overall confidence is calculated from relevance, logic, and refinement signals. Weights control the influence of each signal in the final confidence score.
Properties of Confidence Policy
| Name | Type | Description |
| FitWeight | decimal | Controls how strongly overall relevance influences confidence. Relevance represents how well the evaluated data aligns with the intended outcome. |
| LogicTrueWeight | decimal | Controls the influence of logic when conditions evaluate as true. Higher values increase confidence when logic validates the result. |
| LogicFalseWeight | decimal | Controls the influence of logic when conditions evaluate as false. Higher values increase the impact of logic rejection on confidence. |
| LogicAlignmentWeight | decimal | Controls how much logic coverage influences confidence. This reflects how much of the evaluated data space logic meaningfully governed. |
| SeparationWeight | decimal | Controls how strongly statistical deviation from the average candidate influences confidence. Higher values increase confidence when a result is significantly better than the overall distribution of evaluated candidates. |
| SeparationUsesAbsoluteDeviation | bool | Determines how separation evaluates statistical deviation among candidates. Enable to reward any extreme deviation from the average (both unusually strong and unusually weak results); disable to reward only candidates that outperform the average. This setting is intended for advanced scoring scenarios. |
404 - Not Found