Files
yao-meta-skill/evals/sample_trigger_report.json
T

38 lines
919 B
JSON

{
"threshold": 0.35,
"threshold_explanation": "Prompts at or above the threshold are treated as trigger matches. Scores near the threshold should be reviewed as boundary cases.",
"false_positives": 2,
"false_negatives": 0,
"precision": 0.667,
"recall": 1.0,
"bucket_stats": {
"should_trigger": {
"total": 4,
"passed": 4,
"pass_rate": 1.0
},
"should_not_trigger": {
"total": 3,
"passed": 3,
"pass_rate": 1.0
},
"near_neighbor": {
"total": 3,
"passed": 1,
"pass_rate": 0.333
}
},
"comparison": {
"baseline_false_positives": 0,
"baseline_false_negatives": 4,
"improved_false_positives": 2,
"improved_false_negatives": 0,
"false_positive_delta": 2,
"false_negative_delta": -4,
"baseline_precision": null,
"improved_precision": 0.667,
"baseline_recall": 0.0,
"improved_recall": 1.0
}
}