Datasets:

junlinw
/

arb-harvey-lab-data-packet

Languages:

English

Size:

Tags:

License:

Dataset card Files Files and versions

xet

Community

Dataset Viewer

The dataset viewer is not available for this subset.

Cannot get the split names for the config 'default' of the dataset.

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 80, in _split_generators
                  raise ValueError(
                  ...<2 lines>...
                  )
              ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
              
              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/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

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.

Harvey LAB data packet

Study material for the harvey-lab benchmark of Applied RSI Bench (ARB): the official texts of the laws, regulations and court opinions that Harvey LAB's rubric lines cite, as plain text. An ARB post-training agent gets one tier, read-only, at /home/agent/workspace/data_packet/ before its model is evaluated.

The packet holds public legal texts only: no LAB scenarios, case files, task instructions or rubrics.

file	docs	unpacked → archive	tree_sha256
`S.tar.gz`	941	92.6 MB → 23.4 MB	`0288dcf78e6e383b76d8564224f13ddcd18722ea220d3367b032cdbfe50e577f`
`L.tar.gz`	21,367	398.5 MB → 85.4 MB	`6877a6bff7c00ea20e25e65c89a06675dcaa8b874ca626c8c3a10537a694ab21`

S: the official texts an agent quoted while matching each cited provision to its source (1,010 authorities, 1,451 cited provisions; 1,266 found in an official source).
L: every cited statute and regulation expanded to its enclosing unit, from the official publisher: the U.S. Code chapter of each cited section, the C.F.R. part, the whole act for UK and Illinois acts, Delaware Code titles 6, 8, 10 and 12, and the chapter or title of each cited state code section; plus each cited court opinion, session law and foreign act as a single document. S ⊆ L.

U.S. Code chapter pages (82.6 MB) and whole-part C.F.R. files (49.7 MB) repeat text that L also holds section by section: about 33% of L's bytes.

Left out: authorities LAB invents for its fictional matters; archived copies of official pages (web.archive.org) that stood in for sites a polite crawler cannot read; sites whose terms need permission (India's MeitY and TRAI, German law translations) or are unclear; commercial case-law sites. California's codes are missing: leginfo.legislature.ca.gov disallows crawling.

Layout

PACKET.json          docs, bytes, tree_sha256, and per file: path, sha256, bytes, url, license, meta
<host>/<slug>.txt    one normalized document per file

tree_sha256 is the sha256 of the sorted lines "{path} {sha256}\n", one per document (PACKET.json excluded).

Use

hf download junlinw/arb-harvey-lab-data-packet L.tar.gz --repo-type dataset --local-dir .
mkdir -p data_packet && tar -xzf L.tar.gz -C data_packet

ARB pins the exact commit it mounts in the benchmark's benchmark.toml; pass --revision <commit> to get the same bytes.

Sources and licenses

Fetched from each publisher's own site with robots.txt honored and per-host rate limits, converted to text (HTML → one block per line; PDF → its text layer), and hashed. These are not official versions; check the publisher before relying on a text. Documents in L by license:

license	docs	hosts
`us-government-work`	15,080	uscode.house.gov (7,643), www.govinfo.gov (7,399), www.cafc.uscourts.gov (12), www.supremecourt.gov (6), cafc.uscourts.gov (5), cdn.ca9.uscourts.gov (3), +10 more
`government-edict`	3,816	www.ilga.gov (1,837), malegislature.gov (500), delcode.delaware.gov (417), app.leg.wa.gov (278), www.ncleg.gov (261), www.revisor.mn.gov (121), +30 more
`open-government-licence-v3`	2,453	www.legislation.gov.uk (2,453)
`reproduction-of-federal-law-order`	6	laws-lois.justice.gc.ca (6)
`eu-reuse-with-attribution`	4	eur-lex.europa.eu (4)
`official-text`	3	www.fedlex.admin.ch (1), ndpc.gov.ng (1), www.planalto.gov.br (1)
`pdl-1.0`	3	www.japaneselawtranslation.go.jp (3)
`cc-by-4.0`	1	api.prod.legislation.gov.au (1)
`public-record`	1	www.sec.gov (1)

government-edict: Statutes, regulations and court opinions are government edicts (Georgia v. Public.Resource.Org, 590 U.S. 255 (2020); Banks v. Manchester, 128 U.S. 244 (1888)).
official-text: Laws and official acts are outside copyright where published: Brazil (Lei 9.610/1998 art. 8 IV), Nigeria (Copyright Act 2022 s. 3(b)), Switzerland (URG art. 5).
us-government-work: U.S. federal works have no copyright (17 U.S.C. § 105).
public-record: SEC filings are public records.
Japanese law translations: source Japanese Law Translation Database System (Ministry of Justice, Japan), Public Data License 1.0; the translations are not official texts.
Canadian federal law reproduced under the Reproduction of Federal Law Order (SI/97-5); this is not an official version.
Based on content from the Federal Register of Legislation at the fetch date. For the latest information on Australian Government legislation please go to https://www.legislation.gov.au. Licensed under CC BY 4.0.
Contains public sector information licensed under the Open Government Licence v3.0.

The packagers claim no rights in the documents.

Downloads last month: 44

Total file size:

109 MB