README.md

---
pretty_name: RTEB · FinanceBenchRetrieval
license: apache-2.0
language:
- en
multilinguality:
- monolingual
task_categories:
- text-retrieval
task_ids:
- document-retrieval
tags:
- rteb
- mteb
- retrieval
- finance
configs:
- config_name: corpus
  data_files:
  - split: test
    path: corpus/test-*.parquet
- config_name: qrels
  data_files:
  - split: test
    path: qrels/test-*.parquet
- config_name: queries
  data_files:
  - split: test
    path: queries/test-*.parquet
---
# FinanceBenchRetrieval — RTEB open subset, unified schema

A normalised copy of the dataset behind the `mteb` task **`FinanceBenchRetrieval`**, one of the 17 open tasks in the
[RTEB(beta)](https://huggingface.co/blog/rteb) retrieval benchmark. Same queries, documents and relevance
judgements as the benchmark evaluates — reshaped into one strict schema shared by all 17.

|  |  |
|---|---|
| Source | [`embedding-benchmark/FinanceBench`](https://huggingface.co/datasets/embedding-benchmark/FinanceBench/tree/e68478442112cae36b70a216f52cc2777acf0a7e) @ `e68478442112` (the revision pinned in `mteb`) |
| Domain · languages | finance · eng |
| Queries / documents / qrels | 150 / 145 / 150 |
| Qrels per query | min 1 · mean 1.0 · max 1 |
| Score values | 1 ×150 |
| Layout | `queries` · `corpus` · `qrels`, split `test` |
| License | `apache-2.0` |

## Schema

| config | columns | rules |
|---|---|---|
| `queries` | `id: string`, `text: string` | ids unique and non-empty; every query has ≥ 1 qrel |
| `corpus` | `id: string`, `title: string`, `text: string` | `title` is always present (`""` when the source has none) |
| `qrels` | `query-id: string`, `corpus-id: string`, `score: int32` | referential integrity to both tables; no duplicate pairs; no floats |

Files are Parquet, sorted by id, zstd-compressed, sharded at 500 MB. Every rule above is enforced by a validator
before publishing; `provenance.json` records the source file hashes, what changed, and the output file hashes.

## What changed from the source

- added a `title` column filled with `""` (the source has none)
- cast `qrels.score` double -> int32
- removed NUL bytes from text (prose dataset)
- renamed source splits (`queries`←`queries/queries`, `corpus`←`corpus/corpus`, `qrels`←`default/test`) to `test`

## Load it

```python
from datasets import load_dataset
queries = load_dataset("Hyukkyu/rteb-FinanceBenchRetrieval", "queries", split="test")
corpus  = load_dataset("Hyukkyu/rteb-FinanceBenchRetrieval", "corpus", split="test")
qrels   = load_dataset("Hyukkyu/rteb-FinanceBenchRetrieval", "qrels", split="test")
```

## License and attribution

The data is redistributed under the source's terms — `apache-2.0`. All credit belongs to the
original authors; see the source repository and the references in `mteb`'s task metadata (https://huggingface.co/datasets/embedding-benchmark/FinanceBench).
This repository is an independent repackaging and is not affiliated with the RTEB or MTEB maintainers.

> **License discrepancy.** The upstream dataset card declares `apache-2.0` while `mteb`'s `TaskMetadata` declares `mit`. This card carries the upstream value as the more conservative choice.