Dataset Viewer
Auto-converted to Parquet Duplicate
pair_id
large_string
left_id
large_string
right_id
large_string
label
int64
id_type
large_string
label_confidence
large_string
difficulty
large_string
categories
large_string
name_sim
float64
split
large_string
p0000cbddf019
r3080b79ba14b
rea977c76dbc5
0
uei
high
easy
state_differs
68.6
train
p000307369f6b
r26dc62d3c832
r24ec173d6d38
0
state_filing
high
hard
same_key;form_conflict;state_differs
100
train
p0003ad1b40cb
r49c14553642d
r99b8ab87410b
0
uei
high
easy
state_differs
65.5
train
p000b74831a26
r9f6cfdc03e23
r7eddb5041257
0
uei
high
easy
form_conflict
40
train
p000ba2879075
r2bd07c2fc9bd
r0fe77ec26cd3
1
uei
high
hard
cross_source;dba_alias
32
train
p000dc082692b
r7d6a2a132dd5
radc6d2c35253
1
uei
high
hard
cross_source;form_conflict
33.3
train
p000ef021347b
r6fd1ce365c4c
rb0b2e4a3b24b
1
uei
high
hard
cross_source
65.4
train
p000f641e5968
rd65729f8f563
r36c17d9abc8e
0
uei
high
medium
state_differs
88.2
train
p0013bc5b0e46
r5405a869b868
r928340cbf034
0
state_filing
high
hard
same_key;form_conflict
100
train
p0013bd215358
r54514e5793c1
r1f223a6a3a1a
1
uei
high
easy
cross_source;same_key
100
train
p00184230597f
recc9091f299b
r8c0f57ebc435
0
ccn
medium
medium
state_differs
76.5
train
p001ad5ea564a
rfd2f22924498
r39fc7bc7572a
1
frn
medium
medium
81.1
train
p001b78b88fd8
ra3711d42388f
r9c82d7272fd9
0
state_filing
high
hard
same_key;form_conflict;state_differs
100
train
p001d645f84da
r6b3e2729d34f
r689cb034268e
1
uei
high
medium
cross_source;dba_alias
85.7
train
p001ecfd61ea3
rafd68ea8e8f9
rd41b01499606
0
ccn
medium
medium
cross_source
82.4
train
p002181f2b824
r6c1228078f68
r6d99b8642812
0
uei
high
easy
state_differs
43.9
train
p00298a1bef00
r536741808e37
r0ec10087c1bf
0
state_filing
medium
hard
same_key;form_conflict
100
train
p002bbe1fe798
r002712059429
rbe562d2e9ceb
1
uei
high
easy
cross_source;same_key;form_conflict
100
train
p002c454742bf
r1cd795001893
r2d7d9226590c
1
frn
medium
easy
same_key
100
train
p002d5474234b
raeab0c11058a
r26b410c42071
1
uei
high
easy
cross_source;same_key
100
train
p002d58e5b021
rb63e0e3429c5
r546d2abbac1e
1
frn
medium
easy
95.1
train
p002d9be3044f
r8c5933c5d580
r5dae50dfe512
0
uei
high
easy
state_differs
53.3
train
p002f902d2194
r066ccef2f51a
re2963903344a
1
uei
high
easy
cross_source;same_key;dba_alias
100
train
p002f9b4ba3cc
r33c5b5d74fb8
rac053cbca0ea
0
state_filing
medium
hard
same_key;form_conflict
100
train
p003480c95f55
r7c92e0394ebb
raadf352ea547
0
cik
high
easy
44.4
train
p0036ebba2aa1
rb5bd6d61c103
rbd17bbab4e43
0
uei
high
medium
70.6
train
p0036eea07dbc
rb123228d9611
rb276761a5b7e
1
uei
high
easy
cross_source;same_key;form_conflict
100
train
p003728cdfd98
r24bc77075e77
ra58ee26fcbe7
0
state_filing
high
hard
same_key;form_conflict
100
train
p003764ad5459
r975767ca0d6d
rade5bb24a675
0
uei
high
medium
state_differs
76.6
train
p0039283e31b1
r887ee3265e45
rdf9db31bc6be
0
state_filing
high
hard
same_key;form_conflict
100
train
p003e2d6217c2
r3d3ba5315d6c
r1e35aa4fce35
1
frn
medium
easy
same_key
100
train
p0040b06e89c5
r3881b4b74dce
rdbb29867de96
1
ccn
high
hard
cross_source
60.6
train
p004162139660
r411342314e62
r35e0324fa8ab
0
uei
high
easy
state_differs
58.8
train
p004248972a30
rf56da2ad8a0d
r8cef7ed443cb
0
state_filing
medium
hard
same_key;form_conflict
100
train
p0043934d1279
rdab8d1ac4d69
r8012b6a7d450
0
uei
high
easy
state_differs
43.5
train
p0044ca3892f0
rcbf185d1f274
r5964972bc8dd
0
uei
high
easy
form_conflict;state_differs
60.6
train
p00450c38dbc4
r8b3d73f72986
r6f574201c138
1
uei
high
easy
cross_source;same_key
100
train
p00459df24a02
rf64039acf171
r22ebb74ad7dd
0
uei
high
easy
state_differs
60.5
train
p0045e42cfb8d
r6909ceb682aa
rd17cd04eaf37
0
uei
high
easy
state_differs
59.1
train
p0048d30635f5
r963aba6efe35
r64d1be4b8f84
1
uei
high
medium
cross_source;dba_alias
76.2
train
p0048f1691b8f
rf90dc853a4bc
r278d8ee7b199
0
uei
high
easy
state_differs
49.2
train
p004b99d1f1c5
ra843061d9bd5
rc16bdfe50dd0
0
state_filing
high
hard
same_key;form_conflict
100
train
p004d3bcc8739
reb3bf1b04284
rb837a79a0046
1
uei
high
medium
cross_source;dba_alias
77.9
train
p004f02897607
r492f9e76625a
r899eb01a1ca4
0
uei
high
hard
same_key;state_differs
100
train
p0053d3406d79
rf173ab772253
r5f94502052b5
1
uei
high
hard
cross_source;form_conflict
30
train
p0053f7d0ebbd
rfd4b21f0ff5e
r68d4dec2b552
1
frn
medium
easy
same_key
100
train
p0058efd50892
r28f741facd26
r4a1afca45e89
0
uei
high
easy
state_differs
51.4
train
p00620b48a5ea
r3080b79ba14b
rfc5061c5d37d
0
uei
high
easy
state_differs
69.2
train
p0064a85476f4
rca90b9c2bb02
radcb1382f010
0
uei
high
hard
cross_source;same_key;form_conflict;state_differs
100
train
p0068a7446e0a
r4a96470338ed
r4cd686c6d5af
1
uei
high
medium
dba_alias
71.4
train
p006ca290ecba
rcb394aa36d27
r6e986d8324be
0
state_filing
medium
hard
same_key;form_conflict
100
train
p006e911bc196
ref103401fe93
r882300a8b1b0
0
state_filing
high
hard
same_key;form_conflict
100
train
p006fdcf616b0
re787712ad3ce
r61f5c1169a9d
0
state_filing
high
hard
same_key;form_conflict
100
train
p00743c393048
rcfea683ca774
r7cb29aee7b10
0
uei
high
medium
state_differs
76.9
train
p00749068ec68
r436c654391cb
r086303a64085
0
state_filing
high
hard
same_key;form_conflict
100
train
p00750cc0c699
r425bda1b64d0
r17ee4e71353e
1
uei
high
easy
cross_source;same_key
100
train
p007783350fed
rabc3ca19919b
rcf0bb937d46f
1
ccn
high
medium
cross_source
88.5
train
p0079214c200d
r8196308aacae
r71910155cdc5
0
uei
high
hard
same_key;state_differs
100
train
p0079497e93dc
r794c9ad74959
r9778f9bf5fdd
0
state_filing
high
hard
same_key;form_conflict
100
train
p007b48c3a4ef
r575368ba3656
rac43073d4bef
0
state_filing
medium
hard
same_key
100
train
p007b88d3d14a
r4a3d6663e9ce
r26c6bd7cc3d0
0
uei
high
medium
cross_source
70
train
p007bac589ac3
r4dc5162b9e2f
r4bf540d5f991
1
frn
medium
medium
81.1
train
p007c3eccaf33
rd8793f17b905
r6da9dda23154
1
uei
high
easy
cross_source
92.1
train
p007ee78b346b
r987420416aae
r96c782d60c6d
0
uei
high
medium
state_differs
73.3
train
p007f8cfc7784
rcfe29b4f0a2f
r759f4ecbbb24
1
uei
high
easy
same_key;dba_alias
100
train
p0080d16ba2ab
rd2e27af0bcc2
re2e24306ee2c
1
ccn
high
easy
cross_source;same_key
100
train
p0080d4886696
r0a5be4b98669
r15effe54d5d6
1
uei
high
easy
cross_source;same_key
100
train
p00838385b910
ra4187fea6bea
r21e9ad87dbb0
0
uei
high
easy
form_conflict;state_differs
54.9
train
p0084347c5700
r3f02accd13df
r7df7b81f7b97
1
cik
high
easy
cross_source;same_key
100
train
p0084ffefbcdb
rc3e5333d6443
r290b6225a236
0
uei
high
medium
form_conflict;state_differs
85.1
train
p00856202c493
r62109a3c35d4
r1e5d3e0215a3
0
state_filing
high
hard
same_key
100
train
p0085e11b897c
r6cf825929f07
r8251a602b15f
0
uei
high
easy
state_differs
37.8
train
p00872d0d56ef
rc934d20ea42e
rbf24c98c883b
0
uei
high
easy
47.4
train
p0088b0571a7b
r435483791953
rad61440caecc
1
uei
high
hard
dba_alias
64
train
p008f32ca8c22
r9554fc25a80e
ra88523668c47
0
uei
high
easy
state_differs
48.9
train
p008fe682413d
r20a1ec45a6cf
r27811fdbabca
1
cik
high
easy
cross_source;same_key
100
train
p00903382995c
r85853d880f37
r19fe61169c0d
0
uei
high
easy
state_differs
45.3
train
p0092d46ba03a
r234803052541
r74354282bdec
0
uei
high
easy
state_differs
37.2
train
p009b315b0f5c
r662d7d309bd7
r12179bdb9698
1
uei
high
easy
dba_alias
90.8
train
p009d0e4e19ca
r3a35748769ea
r6606c6b88732
0
uei
high
hard
same_key;state_differs
100
train
p009f1c7379f2
rcd5cd704015d
r162842f7f395
1
uei
high
easy
cross_source;same_key;dba_alias
100
train
p009f2245b8bc
r4caae62c14ea
r01374db49738
0
ccn
medium
medium
state_differs
75.6
train
p009f46b08a18
r1fc6935d7210
rf9e2c8186722
0
uei
high
easy
form_conflict
52.2
train
p009f97b53087
rbdef8e3270d2
rf9bc733e80e6
0
state_filing
high
hard
same_key;form_conflict
100
train
p009fb8bcded9
r54f57c7624de
r260c20c6039b
1
uei
high
medium
cross_source
80
train
p00a2bd91c769
r79aa61db73ea
r6c97f693ed24
0
uei
high
medium
cross_source;state_differs
71
train
p00a35fc10173
r9d8d76ec23e5
rd11856fa2bfd
0
state_filing
high
hard
same_key;form_conflict
100
train
p00a465f6dade
ra2c74e66e0a4
rd69b7c45335c
0
uei
high
easy
form_conflict;state_differs
23.8
train
p00a67a60cb25
rb111237539c5
rc738cc6ba079
1
uei
high
easy
cross_source;same_key
100
train
p00a871ac70f2
r45f5c32fabb0
r3bf720c513e6
0
uei
high
easy
form_conflict;state_differs
51.6
train
p00a8e0fab0cc
r7a16b8af5cc7
r2cd88eb87485
1
uei
high
hard
cross_source;dba_alias
24.2
train
p00aad7627bf5
rd7dd42edd5ce
rfcdb25f0f78d
1
uei
high
easy
cross_source;same_key;dba_alias
100
train
p00ab6f28eef5
r45b7c5043d6b
r74349f95d1ea
0
state_filing
medium
hard
same_key;form_conflict
100
train
p00abab2f306f
r7c9d37dd8a24
r2c8b7bbcbff2
1
uei
high
easy
cross_source;same_key
100
train
p00b1a5c7cd41
r201d37aaab76
r87a00f392b6c
0
cik
high
medium
71.4
train
p00b254375650
r21934cc3078f
r243e8019cfc1
0
state_filing
medium
hard
same_key;form_conflict
100
train
p00b3507d6698
r8de356375e04
r1ee2bad46296
0
uei
high
medium
84.6
train
p00b394615ffc
r831982199197
r5f58a520e681
0
uei
high
medium
state_differs
73.5
train
p00b40b166fad
r655006524127
r18abbf646e36
0
uei
high
easy
60.7
train
p00b4f86e8277
r651cff74c9cb
r2265867c3404
0
state_filing
high
hard
same_key;state_differs
100
train
End of preview. Expand in Data Studio

USORG-Bench

A US organization matching benchmark whose labels come from government identifiers. Two public records match when they carry the same SAM.gov UEI, SEC CIK, CMS certification number, FCC registration number, USAC billed-entity or service-provider number, or state business filing number, and do not match when they carry different values of the same identifier.

  • 62,497 labelled pairs (26,149 matches, 36,348 non-matches) over 96,642 records and 71,783 distinct identifiers (one organization can hold several), from 15 public sources: USAspending (FPDS contracts, recipient profiles, DoD subawards), SEC EDGAR, CMS hospital data, FCC ULS and LMS, USAC E-Rate, and the Colorado, Connecticut, New York and Oregon business registries.
  • Organizations only: sole proprietorships and person names are removed.
  • Identifier values are masked (entity_id is a keyed HMAC); names and business addresses are as the sources print them.
  • train / dev / test splits disjoint by identifier (31,199 / 6,123 / 25,175 pairs). In datasets.load_dataset they are train, validation and test; records of a split are the ones its pairs reference.
  • difficulty strata (renames and reworded names are hard matches; near-identical names of different organizations are hard non-matches) and a label_confidence column.
  • A 300-pair audit (audit/audit_300.csv).

Code, loader, metrics, baselines and the build scripts: https://github.com/Mattbusel/usorg-bench (pip install usorg-bench once published; until then pip install git+https://github.com/Mattbusel/usorg-bench).

Leaderboard rule: allowed inputs

A reported result may read only name, address, city, state and zip5 (and the ids as opaque keys). source, id_type, entity_id, name_kind and the pair columns (difficulty, categories, name_sim, label_confidence) follow from how the pairs were built and predict the label without reading a name. Measured on the test split with a lookup table fit on train and the threshold picked on dev:

key looked up test F1 test AP
id_type + the two records' source 0.927 0.908
the same + state equal + names identical 0.973 0.980
difficulty + categories 0.923 0.976
state equal (an allowed field) 0.800 0.680

The last line is a property of the sampling: the build drops same-state near-name non-matches, so "same state" alone is worth F1 0.800. Read record-level results against that, not against the name-only all-match F1 of 0.588. Also by construction: every pair whose names are identical after case and punctuation folding is a non-match (2,113 pairs, all but one in different states). Details: docs/metrics.md.

Files

path contents
data/records.{csv,parquet} one row per source record
data/pairs.{csv,parquet} labelled pairs with split, difficulty, confidence, categories
data/pairs-{train,dev,test}.parquet, data/records-{train,dev,test}.parquet the same rows, one file per split (since v1.0.1)
fixture/ the 5,000-pair subset bundled with the Python package (CSV, and Parquet per split)
audit/audit_300.csv 300 audited test pairs with verdicts
baselines/ baseline scores (pair_id,score) and results.json
exports/name_bench/ 300 rows in nomenklatura's contrib/name_bench/cases.csv format
exports/senzing/, exports/senzing_full/ Senzing V4 JSONL records and truth-set key
exports/zingg/ Zingg example: data, training pairs, config.json
build_stats.json every count and drop from the build

Column definitions: docs/format.md. Datasheet: docs/datasheet.md.

Quick start

from datasets import load_dataset
pairs = load_dataset("tensorust/usorg-bench", "pairs", revision="v1.0.1")      # train / validation / test
records = load_dataset("tensorust/usorg-bench", "records", revision="v1.0.1")

or with pandas:

import pandas as pd
base = "hf://datasets/tensorust/usorg-bench@v1.0.1/data"
test = pd.read_parquet(f"{base}/pairs-test.parquet")
records = pd.read_parquet(f"{base}/records.parquet").set_index("record_id")
allowed = ["name", "address", "city", "state", "zip5"]
left, right = records.loc[test.left_id, allowed], records.loc[test.right_id, allowed]

or with the package:

import usorg_bench
data = usorg_bench.load("full")          # downloads and checks this revision
for pair_id, a, b in data.select("test").iter_feature_pairs():   # allowed fields only
    ...

Score each pair from 0 to 1, pick the threshold on dev, report F1 and average precision on test (usorg-bench eval scores.csv does this; include the dev scores in the file).

Baselines (test split, threshold chosen on dev)

baseline fields read test F1 precision recall average precision dev threshold
splink name + city, state, ZIP 0.876 0.861 0.891 0.864 0.0024
logicv2_full name + address 0.624 0.561 0.703 0.628 0.6989
logicv2 name 0.588 0.416 1.000 0.510 0.0000
rapidfuzz name 0.588 0.416 1.000 0.498 0.0000
reference: state equal state 0.800 0.696 0.941 0.680 0.6966

Name similarity alone does not separate the classes by design: hard non-matches are organizations with the same or nearly the same name (two towns called Springfield), and hard matches are renamed or reworded organizations. A method that reads the address fields can tell many of the same-name pairs apart, but part of that is the same-state sampling effect above.

Audit

300 random test pairs (50 per label and difficulty) were read one by one by the build agent (Claude, an AI model, not a human annotator): 287 agree, 5 disagree, 8 unclear (1.7% and 2.7%). The release filters written after that review remove the 5 error patterns (503 pairs in all); of the 293 audited pairs left in the release, 285 agree and 8 are unclear. Details in the datasheet.

Revisions

  • v1.0 (2026-11-17): first release.
  • v1.0.1 (2026-11-17): the same rows; adds one Parquet file per split and maps them to train / validation / test (in v1.0 every pair sat in one train split of the viewer), fixes the fixture_pairs config, and adds the leaderboard rule above. usorg-bench 0.1.1 loads this revision.

License and citation

CC BY 4.0. The underlying federal records are public domain; state registry records are public records.

@misc{usorgbench2026,
  title  = {USORG-Bench: a US organization matching benchmark labelled by government identifiers},
  author = {{Tensorust LLC}},
  year   = {2026},
  url    = {https://huggingface.co/datasets/tensorust/usorg-bench}
}
Downloads last month
124