Skip to main content

Module conditioning_eval

Module conditioning_eval 

Source
Expand description

Cold, bounded evaluation API for prompt precision (Prompt Precision P5).

Re-exports§

pub use corpus::SplitType;
pub use corpus::TaskItem;
pub use corpus::TaskManifest;
pub use report::EvaluationReport;
pub use report::PairedComparison;
pub use scoring::evaluate_task_output;
pub use scoring::TaskScore;

Modules§

corpus
Versioned task manifests and split enforcement for held-out evaluation.
report
Paired comparison reporting for baseline vs compiled conditioning.
scoring
Task output scoring and validation against ground truth.