Dataset Preview
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
retrieval_model: string
n_docs: int64
depth: int64
configs: list<item: list<item: string>>
child 0, item: list<item: string>
child 0, item: string
id: string
title: string
sentences: list<item: string>
child 0, item: string
to
{'id': Value('string'), 'title': Value('string'), 'sentences': List(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
retrieval_model: string
n_docs: int64
depth: int64
configs: list<item: list<item: string>>
child 0, item: list<item: string>
child 0, item: string
id: string
title: string
sentences: list<item: string>
child 0, item: string
to
{'id': Value('string'), 'title': Value('string'), 'sentences': List(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 datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
id string | title string | sentences list |
|---|---|---|
doc_0 | Trade winds | [
"Trade winds - wikipedia Trade winds Jump to : navigation , search This article is about the weather phenomenon .",
"For other uses , see Tradewind .",
"The westerlies ( blue arrows ) and trade winds ( yellow and brown arrows ) The trade winds are the prevailing pattern of easterly surface winds found in the tr... |
doc_1 | List of Nobel laureates in Physics | [
"List of Nobel laureates in Physics - wikipedia List of Nobel laureates in Physics Jump to : navigation , search Front side ( obverse ) of the Nobel Prize Medal for Physics presented to Edward Victor Appleton in 1947 The Nobel Prize in Physics ( Swedish : Nobelpriset i fysik ) is awarded annually by the Royal Swedi... |
doc_2 | Deadpool 2 | [
"Deadpool 2 - wikipedia Deadpool 2 Jump to : navigation , search Deadpool 2 Teaser poster Directed by David Leitch Produced by Simon Kinberg Ryan Reynolds Lauren Shuler Donner Written by Rhett Reese Paul Wernick Ryan Reynolds Based on Deadpool by Rob Liefeld Fabian Nicieza Starring Ryan Reynolds Josh Brolin Morena ... |
doc_3 | Fortnite | [
"Fortnite - wikipedia Fortnite Jump to : navigation , search For the battle royale version of the game , see Fortnite Battle Royale .",
"Fortnite European box art Developer ( s ) Epic Games People Can Fly Publisher ( s ) Epic Games Director ( s ) Darren Sugg Artist ( s ) Pete Ellis Composer ( s ) Rom Di Prisco En... |
doc_4 | Shortwave radio | [
"Shortwave radio - wikipedia Shortwave radio Jump to : navigation , search `` Shortwave '' redirects here .",
"For other uses , see Shortwave ( disambiguation ) .",
"Grundig Satellit 400 solid - state , digital shortwave receiver , circa 1986 Shortwave radio is radio transmission using shortwave radio frequenci... |
doc_5 | Geography of Nigeria | ["Geography of Nigeria - wikipedia Geography of Nigeria Jump to : navigation , search Location of Ni(...TRUNCATED) |
doc_6 | Health (gaming) | ["Health ( Gaming ) - wikipedia Health ( Gaming ) Jump to : navigation , search A health bar , a pos(...TRUNCATED) |
doc_7 | Cyrus Cylinder | ["Cyrus Cylinder - wikipedia Cyrus Cylinder Jump to : navigation , search Cyrus Cylinder The Cyrus C(...TRUNCATED) |
doc_8 | Governor-General of India | ["Governor - General of India - Wikipedia Governor - General of India Jump to : navigation , search (...TRUNCATED) |
doc_9 | Reading F.C. | ["Reading FC - Wikipedia Reading FC Jump to : navigation , search Not to be confused with Reading F.(...TRUNCATED) |
End of preview.
RAG chunking demo assets (Natural Questions bench + prebuilt FAISS indices)
Prebuilt, read-only assets that let an interactive retrieval demo start instantly without embedding a corpus at boot.
Contents
| path | what it is |
|---|---|
data/nq/large_n1000/docs.jsonl |
1000 Natural Questions Wikipedia documents, pre-split into sentences (id, title, sentences) |
data/nq/large_n1000/questions.jsonl |
1032 NQ validation questions with short answers and gold document titles |
data/nq/large_n1000/indices/demo/index_fixed.{faiss,json} |
FAISS index over the corpus chunked at a fixed 15 sentences / overlap 0 — 19507 chunks, mean 14.65 sentences |
data/nq/large_n1000/indices/demo/index_bilstm.{faiss,json} |
FAISS index over the same corpus chunked by a learned BiLSTM boundary model at target size 15 — 19306 chunks, mean 14.80 sentences |
Both indices are IndexFlatIP over BAAI/bge-base-en-v1.5 embeddings,
L2-normalised so inner product equals cosine similarity. The .json sidecar
holds each chunk's text, sentence count and source document id, so a retrieved
chunk can be traced back to its gold document.
The directory layout mirrors the study's data root, so a consumer can point its data-root environment variable at a snapshot of this repo and resolve every path unchanged.
Provenance
These are the exact indices behind the study's large-scale evaluation — their chunk counts and mean chunk sizes match the archived results row for row. Study, code and full write-up: https://github.com/sfczaa/rag-chunking-ablation
Licensing
Metadata and index structures: MIT. The underlying Wikipedia passage text comes from Natural Questions and remains under CC BY-SA 3.0.
- Downloads last month
- 7