Get a model comparison
Hand off to an LLM
Get a comparison's metrics.
Returns each entry's metric bundle (aggregate + per-field accuracy, latency,
dataset baseline) and its lift against the baseline entry. matchMode
(strict | normalized | semantic) and orderMatching control how leaf
values and arrays are judged equal, exactly as for lab metrics.
Authorization
API Key x-api-key<token>
Authenticate using API Key in request header
In: header
Path Parameters
comparisonID*string
The comparisonID returned by POST /v3/model-comparisons.
Query Parameters
matchMode?string
strict (default), normalized, or semantic.
orderMatching?boolean
Score array elements in order instead of as sets.
includeRowResults?boolean
Include per-row, per-field comparison detail (rowResults) on each entry,
re-matched under matchMode. Powers mislabel detection. Off by default — it
can be large.
Response Body
application/json
application/json
curl -X GET "https://api.bem.ai/v3/model-comparisons/string"{
"comparisonID": "string",
"name": "string",
"status": "pending",
"entries": [
{
"label": "string",
"functionName": "string",
"functionVersionNum": 0,
"scoreRunID": "string",
"status": "pending",
"isBaseline": true,
"coverage": {
"completed": 0,
"total": 0
},
"metrics": {
"aggregateMetrics": {
"accuracy": 0.1,
"precision": 0.1,
"recall": 0.1,
"f1Score": 0.1,
"tp": 0,
"fp": 0,
"fn": 0,
"tn": 0
},
"fieldMetrics": [
null
]
},
"latencyPercentiles": {
"latencyP50": 0.1,
"latencyP90": 0.1,
"latencyP95": 0.1
},
"baselineMetrics": {
"aggregateMetrics": {
"accuracy": 0.1,
"precision": 0.1,
"recall": 0.1,
"f1Score": 0.1,
"tp": 0,
"fp": 0,
"fn": 0,
"tn": 0
},
"fieldMetrics": [
null
]
},
"lift": {
"accuracy": {
"baselineValue": 0.1,
"comparisonValue": 0.1,
"difference": 0.1,
"liftPercent": 0.1
},
"precision": {
"baselineValue": 0.1,
"comparisonValue": 0.1,
"difference": 0.1,
"liftPercent": 0.1
},
"recall": {
"baselineValue": 0.1,
"comparisonValue": 0.1,
"difference": 0.1,
"liftPercent": 0.1
},
"f1Score": {
"baselineValue": 0.1,
"comparisonValue": 0.1,
"difference": 0.1,
"liftPercent": 0.1
}
},
"rowResults": [
{
"rowKey": "string",
"fields": [
{
"path": "string",
"category": "string",
"expected": null,
"actual": null
}
]
}
]
}
]
}{
"message": "string",
"code": 0,
"details": {}
}See also
- System overview — evaluating extraction quality