Rescore Evaluation

Rescore an existing evaluation run with different scorer(s).

Applies the provided scorer(s) to the predictions from source_evaluation_run_id and returns a new evaluation_run_id. Original prediction call IDs are preserved.

POST /evaluations/rescore

Rescore Evaluation

cURL

curl --request POST \
  --url https://api.example.com/evaluations/rescore \
  --header 'Content-Type: application/json' \
  --data '
{
  "source_evaluation_run_id": "<string>",
  "scorer_refs": [
    "<string>"
  ],
  "project_id": "<string>",
  "wb_user_id": "<string>"
}
'

200

{
  "call_id": "<string>",
  "evaluation_run_id": "<string>"
}

Body

application/json

Full rescore request including server-set fields.

source_evaluation_run_id · string · required

The evaluation run whose predictions will be rescored

scorer_refs · string[] · required

Scorer references (weave:// URIs) to apply; must be non-empty

Minimum array length: 1

project_id · string · required

wb_user_id · string | null

Do not set directly. Server will automatically populate this field.

Response

200

application/json

Successful Response

Response for a rescore request.

call_id · string · required

Call ID for /evaluations/status polling

evaluation_run_id · string · required

The newly created EvaluationRun ID