---
configs:
- config_name: bm25
  data_files:
  - split: ar
    path: bm25/ar-*
  - split: bn
    path: bm25/bn-*
  - split: de
    path: bm25/de-*
  - split: en
    path: bm25/en-*
  - split: es
    path: bm25/es-*
  - split: fa
    path: bm25/fa-*
  - split: fi
    path: bm25/fi-*
  - split: fr
    path: bm25/fr-*
  - split: hi
    path: bm25/hi-*
  - split: id
    path: bm25/id-*
  - split: ja
    path: bm25/ja-*
  - split: ko
    path: bm25/ko-*
  - split: ru
    path: bm25/ru-*
  - split: sw
    path: bm25/sw-*
  - split: te
    path: bm25/te-*
  - split: th
    path: bm25/th-*
  - split: yo
    path: bm25/yo-*
  - split: zh
    path: bm25/zh-*
- config_name: corpus
  data_files:
  - split: ar
    path: corpus/ar-*
  - split: bn
    path: corpus/bn-*
  - split: de
    path: corpus/de-*
  - split: en
    path: corpus/en-*
  - split: es
    path: corpus/es-*
  - split: fa
    path: corpus/fa-*
  - split: fi
    path: corpus/fi-*
  - split: fr
    path: corpus/fr-*
  - split: hi
    path: corpus/hi-*
  - split: id
    path: corpus/id-*
  - split: ja
    path: corpus/ja-*
  - split: ko
    path: corpus/ko-*
  - split: ru
    path: corpus/ru-*
  - split: sw
    path: corpus/sw-*
  - split: te
    path: corpus/te-*
  - split: th
    path: corpus/th-*
  - split: yo
    path: corpus/yo-*
  - split: zh
    path: corpus/zh-*
- config_name: qrels
  data_files:
  - split: ar
    path: qrels/ar-*
  - split: bn
    path: qrels/bn-*
  - split: de
    path: qrels/de-*
  - split: en
    path: qrels/en-*
  - split: es
    path: qrels/es-*
  - split: fa
    path: qrels/fa-*
  - split: fi
    path: qrels/fi-*
  - split: fr
    path: qrels/fr-*
  - split: hi
    path: qrels/hi-*
  - split: id
    path: qrels/id-*
  - split: ja
    path: qrels/ja-*
  - split: ko
    path: qrels/ko-*
  - split: ru
    path: qrels/ru-*
  - split: sw
    path: qrels/sw-*
  - split: te
    path: qrels/te-*
  - split: th
    path: qrels/th-*
  - split: yo
    path: qrels/yo-*
  - split: zh
    path: qrels/zh-*
- config_name: queries
  data_files:
  - split: ar
    path: queries/ar-*
  - split: bn
    path: queries/bn-*
  - split: de
    path: queries/de-*
  - split: en
    path: queries/en-*
  - split: es
    path: queries/es-*
  - split: fa
    path: queries/fa-*
  - split: fi
    path: queries/fi-*
  - split: fr
    path: queries/fr-*
  - split: hi
    path: queries/hi-*
  - split: id
    path: queries/id-*
  - split: ja
    path: queries/ja-*
  - split: ko
    path: queries/ko-*
  - split: ru
    path: queries/ru-*
  - split: sw
    path: queries/sw-*
  - split: te
    path: queries/te-*
  - split: th
    path: queries/th-*
  - split: yo
    path: queries/yo-*
  - split: zh
    path: queries/zh-*
license: other
dataset_info:
- config_name: bm25
  features:
  - name: query-id
    dtype: string
  - name: corpus-ids
    list: string
  splits:
  - name: ar
    num_bytes: 248341
    num_examples: 200
  - name: bn
    num_bytes: 233471
    num_examples: 200
  - name: de
    num_bytes: 256575
    num_examples: 200
  - name: en
    num_bytes: 269340
    num_examples: 200
  - name: es
    num_bytes: 255281
    num_examples: 200
  - name: fa
    num_bytes: 249507
    num_examples: 200
  - name: fi
    num_bytes: 242201
    num_examples: 200
  - name: fr
    num_bytes: 259969
    num_examples: 200
  - name: hi
    num_bytes: 240592
    num_examples: 200
  - name: id
    num_bytes: 232601
    num_examples: 200
  - name: ja
    num_bytes: 251987
    num_examples: 200
  - name: ko
    num_bytes: 241907
    num_examples: 200
  - name: ru
    num_bytes: 257759
    num_examples: 200
  - name: sw
    num_bytes: 222259
    num_examples: 200
  - name: te
    num_bytes: 95444
    num_examples: 84
  - name: th
    num_bytes: 238733
    num_examples: 200
  - name: yo
    num_bytes: 132768
    num_examples: 119
  - name: zh
    num_bytes: 247634
    num_examples: 200
  download_size: 740288
  dataset_size: 4176369
- config_name: corpus
  features:
  - name: _id
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: ar
    num_bytes: 2294341
    num_examples: 1854
  - name: bn
    num_bytes: 3347585
    num_examples: 1731
  - name: de
    num_bytes: 1149758
    num_examples: 1748
  - name: en
    num_bytes: 1290954
    num_examples: 1657
  - name: es
    num_bytes: 842327
    num_examples: 1312
  - name: fa
    num_bytes: 1678106
    num_examples: 1858
  - name: fi
    num_bytes: 1268275
    num_examples: 1828
  - name: fr
    num_bytes: 1052962
    num_examples: 1777
  - name: hi
    num_bytes: 2633668
    num_examples: 1748
  - name: id
    num_bytes: 1054118
    num_examples: 1520
  - name: ja
    num_bytes: 1584153
    num_examples: 1846
  - name: ko
    num_bytes: 1673878
    num_examples: 2419
  - name: ru
    num_bytes: 2483294
    num_examples: 1727
  - name: sw
    num_bytes: 877891
    num_examples: 1600
  - name: te
    num_bytes: 1555923
    num_examples: 754
  - name: th
    num_bytes: 3206876
    num_examples: 1897
  - name: yo
    num_bytes: 456903
    num_examples: 921
  - name: zh
    num_bytes: 882899
    num_examples: 1700
  download_size: 14460682
  dataset_size: 29333911
- config_name: qrels
  features:
  - name: query-id
    dtype: string
  - name: corpus-id
    dtype: string
  splits:
  - name: ar
    num_bytes: 3754
    num_examples: 200
  - name: bn
    num_bytes: 3620
    num_examples: 200
  - name: de
    num_bytes: 5041
    num_examples: 200
  - name: en
    num_bytes: 4056
    num_examples: 200
  - name: es
    num_bytes: 5054
    num_examples: 200
  - name: fa
    num_bytes: 4925
    num_examples: 200
  - name: fi
    num_bytes: 3811
    num_examples: 200
  - name: fr
    num_bytes: 5075
    num_examples: 200
  - name: hi
    num_bytes: 4720
    num_examples: 200
  - name: id
    num_bytes: 3699
    num_examples: 200
  - name: ja
    num_bytes: 3839
    num_examples: 200
  - name: ko
    num_bytes: 3777
    num_examples: 200
  - name: ru
    num_bytes: 3864
    num_examples: 200
  - name: sw
    num_bytes: 3550
    num_examples: 200
  - name: te
    num_bytes: 1592
    num_examples: 84
  - name: th
    num_bytes: 3685
    num_examples: 200
  - name: yo
    num_bytes: 2600
    num_examples: 119
  - name: zh
    num_bytes: 4939
    num_examples: 200
  download_size: 79062
  dataset_size: 71601
- config_name: queries
  features:
  - name: _id
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: ar
    num_bytes: 13271
    num_examples: 200
  - name: bn
    num_bytes: 27235
    num_examples: 200
  - name: de
    num_bytes: 12670
    num_examples: 200
  - name: en
    num_bytes: 10242
    num_examples: 200
  - name: es
    num_bytes: 13604
    num_examples: 200
  - name: fa
    num_bytes: 17958
    num_examples: 200
  - name: fi
    num_bytes: 9987
    num_examples: 200
  - name: fr
    num_bytes: 12361
    num_examples: 200
  - name: hi
    num_bytes: 31659
    num_examples: 200
  - name: id
    num_bytes: 9919
    num_examples: 200
  - name: ja
    num_bytes: 12551
    num_examples: 200
  - name: ko
    num_bytes: 13045
    num_examples: 200
  - name: ru
    num_bytes: 18950
    num_examples: 200
  - name: sw
    num_bytes: 9853
    num_examples: 200
  - name: te
    num_bytes: 9661
    num_examples: 84
  - name: th
    num_bytes: 27231
    num_examples: 200
  - name: yo
    num_bytes: 6697
    num_examples: 119
  - name: zh
    num_bytes: 9826
    num_examples: 200
  download_size: 179110
  dataset_size: 266720
---
# NanoMIRACL

This dataset is a Nano-style retrieval dataset. Nano-series evaluation can
be run easily with the [HAKARI Benchmark](https://github.com/hotchpotch/hakari-bench).

NanoMIRACL is derived from hotchpotch/miracl-hf-unified. It follows the Hugging Face
Datasets layout convention used by
[sentence-transformers/NanoBEIR-en](https://huggingface.co/datasets/sentence-transformers/NanoBEIR-en):
each Nano split has separate `corpus`, `queries`, and `qrels` tables, and BM25
candidates are provided separately in a `bm25` table. This layout follows
the NanoBEIR-style evaluation approach summarized in
[NanoBEIR](https://huggingface.co/blog/sionic-ai/eval-sionic-nano-beir).

NanoMIRACL contains multilingual retrieval splits derived from `hotchpotch/miracl-hf-unified`. The `testB` query split is used when available, with a fallback recorded in split metadata when the requested split is not usable.

## Source Links

- [MIRACL unified source dataset](https://huggingface.co/datasets/hotchpotch/miracl-hf-unified)

## Data Layout

This dataset uses four Hugging Face Datasets configs:

- `corpus`: documents with `_id` and `text`
- `queries`: queries with `_id` and `text`
- `qrels`: positive relevance labels with `query-id` and `corpus-id`
- `bm25`: BM25 candidate lists with `query-id` and `corpus-ids`

Each config has the same Nano split names. The exact parquet paths are defined
in the dataset card metadata above. If a regenerated dataset uses a different
schema, config name, path layout, or field name, revise this section before
publishing the README.

## Construction Steps

This dataset was built as follows. If the actual generation procedure differs,
revise this section before publishing the README.

- For each selected query, one positive document is kept and negatives are sampled from that query's source negatives before corpus fill.

1. Use hotchpotch/miracl-hf-unified as the upstream benchmark or dataset family.
2. Load source datasets from `hotchpotch/miracl-hf-unified`.
3. Use `testB` queries when usable, with recorded per-language fallback when necessary.
4. Create one Nano split for each selected source retrieval task.
5. Keep up to 200 eligible queries per Nano split.
6. Include all qrels-positive documents for the selected queries.
7. Fill the corpus from source corpus order up to 10000 documents.
8. Remove exact duplicate document text within each split. If a removed duplicate was referenced by qrels, rewrite qrels to the retained document id.
9. Store corpus text in the generated document `text` field.
10. Generate BM25 top-100 candidates with `per-split auto` tokenization, or the per-split tokenizer shown below.
11. If a qrels-positive document is missing from the raw BM25 result, insert it into the final `bm25` candidate list by replacing a tail non-positive candidate.

## BM25 Subset Policy

The `bm25` config is a candidate subset for first-stage retrieval and reranking.
It is not a separate source dataset. Each row contains one query id and a ranked
list of up to 100 corpus ids.

BM25 candidates are generated from the selected corpus for each split. When a
qrels-positive document is not present in the raw BM25 top-100 results, the
missing positive is forced into the final candidate list by replacing a tail
candidate that is not positive for that query. Candidate ids are kept unique
after replacement.

## Split Mapping

| Nano split | Source task | Source dataset | Queries | Corpus | Qrels |
|---|---|---|---:|---:|---:|
| `ar` | ar_queries | hotchpotch/miracl-hf-unified | 200 | 1854 | 200 |
| `bn` | bn_queries | hotchpotch/miracl-hf-unified | 200 | 1731 | 200 |
| `de` | de_queries | hotchpotch/miracl-hf-unified | 200 | 1748 | 200 |
| `en` | en_queries | hotchpotch/miracl-hf-unified | 200 | 1657 | 200 |
| `es` | es_queries | hotchpotch/miracl-hf-unified | 200 | 1312 | 200 |
| `fa` | fa_queries | hotchpotch/miracl-hf-unified | 200 | 1858 | 200 |
| `fi` | fi_queries | hotchpotch/miracl-hf-unified | 200 | 1828 | 200 |
| `fr` | fr_queries | hotchpotch/miracl-hf-unified | 200 | 1777 | 200 |
| `hi` | hi_queries | hotchpotch/miracl-hf-unified | 200 | 1748 | 200 |
| `id` | id_queries | hotchpotch/miracl-hf-unified | 200 | 1520 | 200 |
| `ja` | ja_queries | hotchpotch/miracl-hf-unified | 200 | 1846 | 200 |
| `ko` | ko_queries | hotchpotch/miracl-hf-unified | 200 | 2419 | 200 |
| `ru` | ru_queries | hotchpotch/miracl-hf-unified | 200 | 1727 | 200 |
| `sw` | sw_queries | hotchpotch/miracl-hf-unified | 200 | 1600 | 200 |
| `te` | te_queries | hotchpotch/miracl-hf-unified | 84 | 754 | 84 |
| `th` | th_queries | hotchpotch/miracl-hf-unified | 200 | 1897 | 200 |
| `yo` | yo_queries | hotchpotch/miracl-hf-unified | 119 | 921 | 119 |
| `zh` | zh_queries | hotchpotch/miracl-hf-unified | 200 | 1700 | 200 |

## BM25 nDCG@10

`nDCG@10` is computed from the included BM25 ranking against the included qrels.

| Nano split | Tokenizer | Forced BM25 positives | BM25 nDCG@10 |
|---|---|---:|---:|
| `ar` | `stemmer:ar/arabic` | 19 | 0.5445 |
| `bn` | `whitespace:bn` | 9 | 0.5103 |
| `de` | `stemmer:de/german` | 39 | 0.3665 |
| `en` | `stemmer:en/english` | 6 | 0.5432 |
| `es` | `stemmer:es/spanish` | 16 | 0.5110 |
| `fa` | `whitespace:fa` | 9 | 0.5337 |
| `fi` | `stemmer:fi/finnish` | 14 | 0.6240 |
| `fr` | `stemmer:fr/french` | 55 | 0.3034 |
| `hi` | `stemmer:hi/hindi` | 6 | 0.5497 |
| `id` | `stemmer:id/indonesian` | 16 | 0.5705 |
| `ja` | `wordseg:ja` | 3 | 0.5956 |
| `ko` | `wordseg:ko` | 5 | 0.5090 |
| `ru` | `stemmer:ru/russian` | 31 | 0.4457 |
| `sw` | `whitespace:sw` | 9 | 0.5782 |
| `te` | `whitespace:te` | 8 | 0.6044 |
| `th` | `wordseg:th` | 7 | 0.6475 |
| `yo` | `whitespace:yo` | 13 | 0.5323 |
| `zh` | `wordseg:zh` | 18 | 0.4466 |

## Skipped Tasks

None.

## License

NanoMIRACL is a derived dataset. Users must comply with the licenses, terms, and
attribution requirements of the upstream datasets listed above.