Hugging Face's logo

Datasets:
SlayerLab
/
plwiki-history-rag-20260920

Tasks:
Text Retrieval
Languages:
Polish
Size:
10K<n<100K
Tags:
wikipedia
history
rag
polish
matura
License:
Dataset card Data Studio Files
xet
 Community
Dataset Preview
Duplicate
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed
Error code:   DatasetGenerationError
Exception:    CastError
Message:      Couldn't cast
id: string
title: string
source: string
text: string
links: list<item: null>
  child 0, item: null
to
{'text': Value('string'), 'source': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              id: string
              title: string
              source: string
              text: string
              links: list<item: null>
                child 0, item: null
              to
              {'text': Value('string'), 'source': Value('string')}
              because column names don't match
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1880, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

text
string
source
string
"II wojna światowa\n\nII wojna światowa – największa i najkrwawsza wojna w dziejach, zaliczana (...TRUNCATED)
https://pl.wikipedia.org/wiki/II_wojna_światowa
"Związek Socjalistycznych Republik Radzieckich\n\nZwiązek Socjalistycznych Republik Radzieckich (s(...TRUNCATED)
https://pl.wikipedia.org/wiki/Związek_Socjalistycznych_Republik_Radzieckich
"III Rzesza\n\nTen artykuł od 2024-10 zawiera treści, przy których brakuje odnośników do źród(...TRUNCATED)
https://pl.wikipedia.org/wiki/III_Rzesza
"I wojna światowa\n\nI wojna światowa (Wielka Wojna) – hegemoniczna wojna światowa, trwająca o(...TRUNCATED)
https://pl.wikipedia.org/wiki/I_wojna_światowa
"Kampania wrześniowa\n\nKampania wrześniowa (także: kampania polska 1939 roku, wojna polska 1939,(...TRUNCATED)
https://pl.wikipedia.org/wiki/Kampania_wrześniowa
"Stan wojenny w Polsce (1981–1983)\n\nStan wojenny w Polsce w latach 1981–1983 – stan nadzwycz(...TRUNCATED)
https://pl.wikipedia.org/wiki/Stan_wojenny_w_Polsce_(1981–1983)
"Mur Berliński\n\nTen artykuł od 2026-04 zawiera treści, przy których brakuje odnośników do ź(...TRUNCATED)
https://pl.wikipedia.org/wiki/Mur_Berliński
"Pakt Ribbentrop-Mołotow\n\nPakt Ribbentrop–Mołotow (też: pakt Hitler–Stalin) – umowa międ(...TRUNCATED)
https://pl.wikipedia.org/wiki/Pakt_Ribbentrop-Mołotow
"Rzeczpospolita Obojga Narodów\n\nRzeczpospolita Obojga Narodów (łac. Res Publica Utriusque Natio(...TRUNCATED)
https://pl.wikipedia.org/wiki/Rzeczpospolita_Obojga_Narodów
"Rozpad ZSRR\n\nRozpad Związku Socjalistycznych Republik Radzieckich – proces rozpadu Związku So(...TRUNCATED)
https://pl.wikipedia.org/wiki/Rozpad_ZSRR
End of preview.

plwiki-history-rag-20260920

The retrieval knowledge base used by team SlayerLab at the Warsaw Model Trainers hackathon (25–27 September 2026, Polish history matura). It is a selection of 62,377 Polish Wikipedia articles relevant to the history curriculum, the same text re-flowed for continued pretraining, and the exact SQLite full-text index (--db) that every registered run of our "Mały, ale wariat" track and the final submission used. Code, recipe and results: github.com/slayerlabs/hackathon (maly-ale-wariat/README.md §1).

Files

File Bytes Records sha256
rag.jsonl 450,977,867 62,377 articles (id, title, source URL, text, links) 1c1a879e1a7556a959b85124478908987060984de1acfbea24c3eae9d24d8d68
plwiki-ABC-c900.sqlite 1,690,390,528 547,482 chunks of 900 characters (overlap 100) in documents + FTS5 documents_fts + common_terms (379) 4accb21efd95db6a5eafb14a353ea540b596b4cc99e8096d51a7e0e8a07d3ec7
cpt.jsonl 305,452,664 33,013 tier-A articles as plain text (continued pretraining) 071596f8e273156c7bb16b072e0a3d69a64a2a2b065623021e964e9b75d77809
stats.json 143 tier counts of the build 6fc89414830bf80efcdb4c855ee73ccf161bbcb5298fbaad31298d97219190fc
selection/corpus.jsonl 454,202,347 62,414 selected articles with tier (A/B/C), selection score, curriculum sections dzialy, url 42fbb7c1d0a37a3a65f0ed60f06277727447b26e1abe044f1a25c9446b7b5723
selection/selected.parquet 2,525,746 the selection table without text 3c94cd78552603138ff4d6e3943bc2b37132ce809ac2cf86938cc705cda47bbc
selection/per_dzial.csv 1,798 articles per curriculum section and tier 575aaf75218491412e715d9000389a8830d54685a3012845ba424ed5435bf15d

SHA256SUMS holds the same hashes; check a download with shasum -a 256 -c SHA256SUMS.

How it was built

Source: the Wikimedia CirrusSearch content dump of Polish Wikipedia dated 2026-09-20 (https://dumps.wikimedia.org/other/cirrussearch/20260920/plwiki_content-20260920-0000{0..7}.json.bz2, 8 shards, 5.35 GB; sha256 of each shard below). Scripts live in the repository above; all defaults unless stated.

  1. matura-wiki-data-fetch-scripts/scripts/01_download.py --cirrus-only --date 20260920 — download and verify the shards.
  2. maly-ale-wariat/data/wiki_pipeline/02_cirrus_to_parquet.py — shards to parquet (meta, text).
  3. maly-ale-wariat/data/wiki_pipeline/04_select.py --seeds seeds_by_dzial.json — curriculum-driven selection against the 2024 history syllabus (podstawa_historia_2024.json, 59 sections, 404 requirements) with seed pages per section; three tiers with history-score thresholds --hist-a 0.5 --hist-b 0.4 --hist-c 0.8, and B/C pages must mention a year ≤ 2015 in the lead (--max-year 2015). Output: selection/ (62,414 articles: A 33,013, B 28,457+, C 907+; see per_dzial.csv).
  4. maly-ale-wariat/data/make_rag_cpt.py --rag-tiers A,B,C --cpt-tiers A --min-chars 300 — rag.jsonl (37 articles under 300 characters dropped) and cpt.jsonl.
  5. python -m matura_harness index rag.jsonl --db plwiki-ABC-c900.sqlite --chunk-size 900 --chunk-overlap 100 (matura-harness, llama.cpp-free) — 547,482 chunks with an FTS5 index, then maly-ale-wariat/data/add_common_terms.py --min-docs 15000 adds the common_terms table so that queries on the 550k-chunk index return in under a second.

The same recipe re-run from the pinned dump reproduces the files byte for byte except for timestamps inside the SQLite header; compare rag.jsonl and the chunk count rather than the sqlite hash if you rebuild.

Cirrus shards (sha256)

Shard Bytes sha256
plwiki_content-20260920-00000.json.bz2 668,628,047 360dc065a30e1fb7ba87a441f3268c113967dfd8b3e9335fcd302492efb098de
plwiki_content-20260920-00001.json.bz2 667,486,569 50351c2de0c6270d745e9a6b24d8b9ea4db4ea7d6964faeb519657adbe5839b8
plwiki_content-20260920-00002.json.bz2 669,587,793 abb980ba80a925e8f953f5fbff72ab6e9239c615f6dcd67cdbab8c3a3cd49a90
plwiki_content-20260920-00003.json.bz2 668,889,710 e84313d4812b9c9fb2a0c6e1f67e219266ba58c5f6aac1d154f09512136c2ed3
plwiki_content-20260920-00004.json.bz2 669,039,404 a55da89cd66c14d97651e062168af4b4d85132ffdf39b2dd9951ea981598c326
plwiki_content-20260920-00005.json.bz2 671,970,336 783c65452ac4a95878c8e043d6f87678bd1631a890a32d72ab9347b8fba88531
plwiki_content-20260920-00006.json.bz2 667,476,266 88683df9a86eb1be6378cdf00ce53937259cc130d3daeb1d27c9f7a07157ca55
plwiki_content-20260920-00007.json.bz2 667,818,540 46459273a47db649621ba09dad7f1cb6d3f441a22c8ebcba78ad9f8a67a5adb2

How to use

With the hackathon harness (matura-harness), pass the index as --db plwiki-ABC-c900.sqlite; the registered harness arguments are in maly-ale-wariat/eval/harness_args_v5.txt. Without the harness, documents has one row per chunk (doc_id, title, source, text) and documents_fts is a standard SQLite FTS5 table over it. rag.jsonl is one JSON object per line and can be re-indexed with any chunker.

Licence and attribution

The text comes from Polish Wikipedia and is licensed CC BY-SA 4.0 by its contributors; this dataset is a derivative work under the same licence. Every article record carries its source / url (https://pl.wikipedia.org/wiki/<title>) for attribution, and the article ids are the Wikipedia page ids. No exam material, marking keys or other third-party content is included.
Downloads last month
182
Total file size:
2.9 GB