doc_id stringlengths 9 80 | doc_path stringlengths 18 89 | gt_path stringlengths 19 90 | schema unknown | suite stringclasses 4
values |
|---|---|---|---|---|
06_10_26_American_Community_Survey_ACS_5_year_estimates_f507236a_sampl | pdfs/06_10_26_American_Community_Survey_ACS_5_year_estimates_f507236a_sampl.pdf | gold/06_10_26_American_Community_Survey_ACS_5_year_estimates_f507236a_sampl.json | [
123,
10,
32,
34,
97,
100,
100,
105,
116,
105,
111,
110,
97,
108,
80,
114,
111,
112,
101,
114,
116,
105,
101,
115,
34,
58,
32,
102,
97,
108,
115,
101,
44,
10,
32,
34,
100,
101,
115,
99,
114,
105,
112,
116,
105,
111,
110,
34... | micro1 |
06_10_26_Budget_vs_actuals_report_cbd029d1_2024_09_SEP_Period_3_MFR_1 | pdfs/06_10_26_Budget_vs_actuals_report_cbd029d1_2024_09_SEP_Period_3_MFR_1.pdf | gold/06_10_26_Budget_vs_actuals_report_cbd029d1_2024_09_SEP_Period_3_MFR_1.json | [
123,
10,
32,
34,
97,
100,
100,
105,
116,
105,
111,
110,
97,
108,
80,
114,
111,
112,
101,
114,
116,
105,
101,
115,
34,
58,
32,
102,
97,
108,
115,
101,
44,
10,
32,
34,
100,
101,
115,
99,
114,
105,
112,
116,
105,
111,
110,
34... | micro1 |
06_10_26_CMS_E_M_Utilization_Report_27e43c5f_cy_2017_evaluation_and_ma | pdfs/06_10_26_CMS_E_M_Utilization_Report_27e43c5f_cy_2017_evaluation_and_ma.pdf | gold/06_10_26_CMS_E_M_Utilization_Report_27e43c5f_cy_2017_evaluation_and_ma.json | [
123,
10,
32,
34,
97,
100,
100,
105,
116,
105,
111,
110,
97,
108,
80,
114,
111,
112,
101,
114,
116,
105,
101,
115,
34,
58,
32,
102,
97,
108,
115,
101,
44,
10,
32,
34,
100,
101,
115,
99,
114,
105,
112,
116,
105,
111,
110,
34... | micro1 |
06_10_26_Federal_Reserve_Z_1_Financial_Accounts_of_the_United_States_e | pdfs/06_10_26_Federal_Reserve_Z_1_Financial_Accounts_of_the_United_States_e.pdf | gold/06_10_26_Federal_Reserve_Z_1_Financial_Accounts_of_the_United_States_e.json | [
123,
10,
32,
34,
97,
100,
100,
105,
116,
105,
111,
110,
97,
108,
80,
114,
111,
112,
101,
114,
116,
105,
101,
115,
34,
58,
32,
102,
97,
108,
115,
101,
44,
10,
32,
34,
100,
101,
115,
99,
114,
105,
112,
116,
105,
111,
110,
34... | micro1 |
06_10_26_Mortality_statistics_and_preventable_mortality_report_d87435b | pdfs/06_10_26_Mortality_statistics_and_preventable_mortality_report_d87435b.pdf | gold/06_10_26_Mortality_statistics_and_preventable_mortality_report_d87435b.json | [
123,
10,
32,
34,
97,
100,
100,
105,
116,
105,
111,
110,
97,
108,
80,
114,
111,
112,
101,
114,
116,
105,
101,
115,
34,
58,
32,
102,
97,
108,
115,
101,
44,
10,
32,
34,
100,
101,
115,
99,
114,
105,
112,
116,
105,
111,
110,
34... | micro1 |
06_19_Bankruptcy_Filings_Statistics | pdfs/06_19_Bankruptcy_Filings_Statistics.pdf | gold/06_19_Bankruptcy_Filings_Statistics.json | [
123,
10,
32,
34,
97,
100,
100,
105,
116,
105,
111,
110,
97,
108,
80,
114,
111,
112,
101,
114,
116,
105,
101,
115,
34,
58,
32,
102,
97,
108,
115,
101,
44,
10,
32,
34,
100,
101,
115,
99,
114,
105,
112,
116,
105,
111,
110,
34... | micro1 |
06_19_Government_zoning_and_land_use_geospatial_datasets | pdfs/06_19_Government_zoning_and_land_use_geospatial_datasets.pdf | gold/06_19_Government_zoning_and_land_use_geospatial_datasets.json | [
123,
10,
32,
34,
97,
100,
100,
105,
116,
105,
111,
110,
97,
108,
80,
114,
111,
112,
101,
114,
116,
105,
101,
115,
34,
58,
32,
102,
97,
108,
115,
101,
44,
10,
32,
34,
100,
101,
115,
99,
114,
105,
112,
116,
105,
111,
110,
34... | micro1 |
06_20_Football_match_results_and_score_tables | pdfs/06_20_Football_match_results_and_score_tables.pdf | gold/06_20_Football_match_results_and_score_tables.json | [
123,
10,
32,
34,
97,
100,
100,
105,
116,
105,
111,
110,
97,
108,
80,
114,
111,
112,
101,
114,
116,
105,
101,
115,
34,
58,
32,
102,
97,
108,
115,
101,
44,
10,
32,
34,
100,
101,
115,
99,
114,
105,
112,
116,
105,
111,
110,
34... | micro1 |
2026-05-13T02-14-01__adversarial_paper_12p__s1 | pdfs/2026-05-13T02-14-01__adversarial_paper_12p__s1.pdf | gold/2026-05-13T02-14-01__adversarial_paper_12p__s1.json | [
123,
10,
32,
34,
116,
121,
112,
101,
34,
58,
32,
34,
111,
98,
106,
101,
99,
116,
34,
44,
10,
32,
34,
116,
105,
116,
108,
101,
34,
58,
32,
34,
65,
99,
97,
100,
101,
109,
105,
99,
32,
80,
97,
112,
101,
114,
32,
72,
101,
... | internal |
2026-05-13T02-14-01__adversarial_paper_12p__s2 | pdfs/2026-05-13T02-14-01__adversarial_paper_12p__s2.pdf | gold/2026-05-13T02-14-01__adversarial_paper_12p__s2.json | [
123,
10,
32,
34,
116,
121,
112,
101,
34,
58,
32,
34,
111,
98,
106,
101,
99,
116,
34,
44,
10,
32,
34,
116,
105,
116,
108,
101,
34,
58,
32,
34,
65,
99,
97,
100,
101,
109,
105,
99,
32,
80,
97,
112,
101,
114,
32,
82,
101,
... | internal |
End of preview. Expand in Data Studio
Omni Extract Bench
We weren’t satisfied with the current benchmarking options for extraction. They were biased, didn’t use realistic data and were hard to audit. Our view is that an extraction benchmark should do two things:
- Help customers choose the right vendor; and
- Give engineers a way to diagnose what’s actually going wrong in a given model.
That’s why we built OmniExtractBench.
OmniExtractBench is a comprehensive structured extraction benchmark, developed by Datalab. It tests how accurately models pull specific values out of documents when given a schema and a PDF. OmniExtractBench includes:
- 620 diverse documents from several vendor benchmarks (Reducto, Extend, LlamaIndex, Datalab) to eliminate bias;
- Clear, consistent and auditable scoring; and
- A wide variety of extraction edge cases tested (scans, dense tables, forms, etc)
Read more in our blog post. Also see the GitHub for how to run the benchmark and use the scorer.
Data
Results
Layout
manifest.parquet one row per document; the schema is in it
pdfs/<doc_id>.pdf the document
gold/<doc_id>.json the ground-truth extraction
licenses/ the upstream licence for each suite
assets/
| column | |
|---|---|
doc_id |
unique; a descriptive name, which the files carry. Some contain spaces |
doc_path |
pdfs/<doc_id>.pdf, relative to the dataset root |
gt_path |
gold/<doc_id>.json, relative to the dataset root |
schema |
the JSON Schema itself, inline as bytes |
suite |
which part of the benchmark it came from |
suite names the part of the benchmark a document came from, and rides through to your
scores, so results can be read per suite as well as overall.
How scoring works
Field matching is deterministic. It works in four stages:
- Normalize the document;
- Flatten the prediction and the ground truth to addresses (the path to a value, such as
line_items[0].qty), each mapped to its scalar value; - Normalize each scalar value, so
"1,240.00"and1240.0are the same number; and - Pair the addresses of the two documents:
- If unambiguous, pair exactly; and
- If ambiguous (the items of an array), Hungarian match on the items' content, recursively for nested arrays.
The full specification is METRIC_SPEC.md on GitHub.
Use
uv pip install 'omni-extract-bench[benchmark]'
How to run the benchmark on this dataset, grade a run, score a prediction on its own, and use the adapters: the omni-extract-bench README on GitHub. This dataset needs omni-extract-bench 1.0.0 or newer.
License
CC BY 4.0. The full texts are in licenses/.
| suite | documents | upstream licence | credit to |
|---|---|---|---|
extractbench |
329 | Apache 2.0 | upstream authors |
internal |
202 | Apache 2.0 | Datalab — synthetic, generated for this benchmark |
longarray |
42 | CC BY 4.0 | Extend AI — cite LongArray-Extract, not this assembly |
micro1 |
47 | MIT | Micro1 |
The scorer itself is separate and Apache 2.0.
Citation
@misc{omni_extract_bench,
title = {Omni Extract Bench},
author = {Datalab},
year = {2026},
url = {https://github.com/datalab-to/omni_extract_bench}
}
- Downloads last month
- 1,704