List Benchmark Datasets

List every benchmark dataset in your workspace with GET /v1/quality/ground-truth. Each dataset holds verified field values that score extraction accuracy.

A benchmark dataset is a collection of manually verified field values that serves as the gold standard for benchmarking extraction accuracy. These datasets and benchmark runs together make up Benchmarks (the /v1/quality namespace). GET /v1/quality/ground-truth lists every dataset in your workspace so you can pick one to benchmark against. This is offline accuracy measurement, not the inline validation checks that gate individual results before delivery.

Do not confuse this resource with the golden samples under [/v1/validation/ground-truth](list-ground-truth): both hold verified values, but they feed different engines. A /v1/quality dataset is schema-scoped, stores one expected_data object per document, and drives repeatable benchmark runs with per-field accuracy scores. A /v1/validation golden sample stores per-field expected values and scores one specific Structuring Run.

Use this endpoint to see all available datasets before creating a benchmark run. A typical workflow is to list datasets, select the one covering the document type you want to evaluate, then pass its id to POST /v1/quality/benchmarks to start a run.

Each dataset includes a name, optional description, user_schema_id (the schema it is scoped to), a document_count, and a links.self URL for the detail endpoint. Datasets are returned in descending creation order with cursor-based pagination: the cursor is keyset-based over (created_at, id), so pages stay stable even while new datasets are being created between requests.

Treat document_count as advisory rather than authoritative: it is maintained by the platform's ground-truth workflow and does not increment when you add entries through POST /v1/quality/ground-truth/:datasetId/entries. To count entries reliably, list them via the [entries endpoint](quality-entries) or read the samples array on the [dataset detail](get-quality-dataset). Benchmark runs are unaffected — documents_total is computed from the live entry count at run creation.

Create separate datasets for different document types or schema versions to track accuracy independently. Pair with the benchmark endpoints to measure extraction accuracy over time: run benchmarks after schema or extraction changes to detect regressions.

• Each dataset contains verified entries mapping documents to expected field values
• Datasets can be scoped to a specific user schema via user_schema_id
• Use datasets as inputs to benchmark runs for per-field accuracy measurement

curl "https://api.talonic.com/v1/quality/ground-truth?limit=20&order=desc" \
  -H "Authorization: Bearer tlnc_..."

Response

{
  "data": [
    {
      "id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
      "name": "Invoice Accuracy Set",
      "description": "Manually verified invoices for Q3 2024",
      "user_schema_id": null,
      "document_count": 50,
      "created_at": "2024-09-01T10:00:00.000Z",
      "links": {
        "self": "/v1/quality/ground-truth/a1b2c3d4-e5f6-7890-abcd-ef1234567890"
      }
    }
  ],
  "pagination": {
    "total": 3,
    "limit": 20,
    "has_more": false,
    "next_cursor": null
  }
}

Errors