Dataset · eval
Arieta Data Legal Eval
Arieta Data Legal Eval is retrieval QA and citation QA built to your specification from the citation graph and version history: your jurisdictions, languages and dates, with point-in-time answers and held-out snapshots.
Tell us your tasks, jurisdictions and dates. We scope the set with you and build it from the graph and the history. We reply within one business day.
Schema
Fields
5 top-level fields. Nested fields are shown with a dotted path. Types are the Parquet physical and logical types; JSONL uses the same names.
| Field | Type | Parquet | Null | Description |
|---|---|---|---|---|
| id | string | BYTE_ARRAY (UTF8) | no | Question id. |
| question | string | BYTE_ARRAY (UTF8) | no | Natural-language question. |
| goldIds | list<string> | LIST<BYTE_ARRAY (UTF8)> | no | Ids of the documents/articles that answer it. |
| asOf | date | INT32 (DATE) | no | Date the answer is valid for. |
| snapshotId | string | BYTE_ARRAY (UTF8) | no | Snapshot the item was derived from. |
Use cases
What it is for
- Retrieval benchmarks with graph-derived gold sets
- Point-in-time QA
- Citation prediction
Related guides
Other datasets
Formats
- Apache Parquet: zstd-compressed, sharded at about 512 MB, one directory per SKU and lane
- JSON Lines (gzip): jsonl.gz, same records and field names as Parquet
Get an eval set built to your spec.
Retrieval QA, citation QA and point-in-time QA from the citation graph and the version history, for your jurisdictions and dates. Tell us what you need to measure.
- We reply within one business day
- Custom cuts in 5–6 business days
- Free sample, no sign-up