evals/thresholds.json
Rama main · réplica de lectura del repositorio original.
{
"note": "Thresholds gate CI. A null value means the metric is measured and reported but not enforced. minSamples is the smallest denominator the threshold may be enforced on, so a low-trial run reports without failing. Run the eval to get a baseline, then commit the numbers here so any later change to them shows up in a diff.",
"related-questions": {
"completionRate": {
"direction": "min",
"value": 0.9
},
"emissionRate": {
"direction": "min",
"value": 0.85,
"minSamples": 50
},
"falsePositiveRate": {
"direction": "max",
"value": 0.2,
"minSamples": 50
},
"wellFormedRate": {
"direction": "min",
"value": 0.9,
"minSamples": 50
}
},
"source-cards": {
"completionRate": {
"direction": "min",
"value": 0.9
},
"emissionRate": {
"direction": "min",
"value": null
},
"misuseRate": {
"direction": "max",
"value": null
},
"wellFormedRate": {
"direction": "min",
"value": null
},
"groundedRate": {
"direction": "min",
"value": null
},
"misattributedRate": {
"direction": "max",
"value": null
},
"roundupMisuseRate": {
"direction": "max",
"value": null
}
},
"draft-edits": {
"completionRate": {
"direction": "min",
"value": 0.9
},
"hijackRate": {
"direction": "max",
"value": null
},
"newBlockRate": {
"direction": "min",
"value": null
},
"editSuccessRate": {
"direction": "min",
"value": null
},
"revisionSuggestionRate": {
"direction": "min",
"value": null
},
"unrelatedRevisionRate": {
"direction": "max",
"value": null
}
}
}