Battam commited on
Commit
aa782f6
·
verified ·
1 Parent(s): 590d908

Release [run release-20261112T140410Z]

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. CHANGELOG.md +21 -0
  2. README.md +18 -18
  3. README.zh.md +18 -18
  4. campaign_files/2026-10-a-spreadsheets-with-answers-2/world-bank-workbooks-odbl/README.md +0 -2
  5. campaign_files/2026-10-a-spreadsheets-with-answers-2/world-bank-workbooks/README.md +0 -2
  6. campaign_files/2026-10-g-solver-instances-2/xbandrade-sokoban/LICENSE +3 -0
  7. campaign_files/2026-10-g-solver-instances-2/xbandrade-sokoban/README.md +23 -0
  8. campaign_files/2026-10-g-solver-instances-2/xcsp3-checking-instances/README.md +0 -8
  9. campaign_files/2026-10-g-solver-instances-2/xhstt-archive/README.md +0 -6
  10. campaign_files/2026-10-g-solver-instances-2/xhstt-high-school-timetabling/README.md +0 -4
  11. campaign_files/2026-10-gt-browser-tests-2/x-wr-timezone-single-events/README.md +0 -4
  12. campaign_files/2026-10-m-issue-trackers-2/webrtc-gerrit-changes/README.md +0 -4
  13. campaign_files/2026-10-m-issue-trackers-2/westcat-gtfs/README.md +0 -4
  14. campaign_files/2026-10-m-issue-trackers-2/whamcloud-jira/README.md +0 -6
  15. campaign_files/2026-10-m-issue-trackers-2/whatwg/README.md +0 -4
  16. campaign_files/2026-10-m-issue-trackers-2/wikimedia-gerrit-changes/README.md +0 -4
  17. campaign_files/2026-10-m-issue-trackers-2/wildfiregames-trac/README.md +0 -8
  18. campaign_files/2026-10-m-issue-trackers-2/xapian-trac/README.md +0 -4
  19. campaign_files/2026-10-md-media-and-documents/when-in-rome-openscore-lieder/README.md +0 -74
  20. campaign_files/2026-10-md-media-and-documents/when-in-rome-variations-and-grounds/README.md +0 -6
  21. campaign_files/2026-10-md-media-and-documents/when-in-rome-wtc1/README.md +0 -4
  22. campaign_files/2026-10-md-media-and-documents/will-hays-citizenship-1919/README.md +0 -4
  23. campaign_files/2026-10-md-media-and-documents/witiko-markdown-table-captions/README.md +0 -4
  24. campaign_files/2026-10-md-media-and-documents/wspr-breqn-testfiles/README.md +0 -4
  25. campaign_files/2026-10-md-media-and-documents/wuffs-test-data/README.md +0 -12
  26. campaign_files/2026-10-md-media-and-documents/xlsx2csv/README.md +0 -4
  27. campaign_files/2026-10-ml-serving-configs-2/wpt-webnn-conformance/README.md +0 -26
  28. campaign_files/2026-10-ml-serving-configs-2/xgrammar-structured-generation-test-expectations/README.md +0 -20
  29. campaign_files/2026-10-ml-serving-configs-2/xilinx-brevitas/README.md +0 -22
  30. campaign_files/2026-10-op-rule-calculations/xbrl-us-dqc-rules-tests/README.md +0 -16
  31. campaign_files/2026-10-s-security-detection-2/welcome-ctf-2025-public-challenges/README.md +0 -12
  32. campaign_files/2026-10-sc-science-reference-data-2/wfa2-lib-utest/README.md +0 -10
  33. campaign_files/2026-10-sw-tests-and-exercises-2/wren-test-suite/README.md +0 -4
  34. campaign_files/2026-10-sw-tests-and-exercises/wpt-url-testdata/README.md +0 -4
  35. campaign_files/2026-10-sw-tests-and-exercises/wycheproof/README.md +0 -4
  36. cards/0xdea-semgrep-rules.md +2 -2
  37. cards/1brc-test-samples.md +2 -2
  38. cards/2dpacklib.md +2 -2
  39. cards/abacus-integrate-tests.md +2 -2
  40. cards/abhishekumarr00-supply-chain-mastery.md +2 -2
  41. cards/academy-bus-nj-gtfs.md +2 -2
  42. cards/accordproject-cicero-template-library.md +2 -2
  43. cards/activity-schema-aggregate-in-between-1.md +2 -2
  44. cards/activity-schema-first-in-between-2.md +2 -2
  45. cards/acwf-verification.md +2 -2
  46. cards/adflib-adf-deleted.md +2 -2
  47. cards/adflib-adf-malformed.md +2 -2
  48. cards/adventure-works-oltp-sample.md +2 -2
  49. cards/agat-tests.md +2 -2
  50. cards/agl-gerrit-changes.md +2 -2
CHANGELOG.md CHANGED
@@ -3501,3 +3501,24 @@ Changes to the content of this index, oldest first. There is one entry for each
3501
  - stackexchange-api-posts: refused
3502
  - uk-government-workbooks: changed pointer group file data/uk-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 1 key(s) of the old version are in no data file of source uk-government-workbooks after the commit (row 22 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3503
  - victoria-government-workbooks: changed pointer group file data/victoria-government-workbooks/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 2 key(s) of the old version are in no data file of source victoria-government-workbooks after the commit (row 6 from batch 2026-10-a-spreadsheets-with-answers, row 7 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3501
  - stackexchange-api-posts: refused
3502
  - uk-government-workbooks: changed pointer group file data/uk-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 1 key(s) of the old version are in no data file of source uk-government-workbooks after the commit (row 22 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3503
  - victoria-government-workbooks: changed pointer group file data/victoria-government-workbooks/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 2 key(s) of the old version are in no data file of source victoria-government-workbooks after the commit (row 6 from batch 2026-10-a-spreadsheets-with-answers, row 7 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3504
+
3505
+ ## 2026-11-12: sources held back
3506
+
3507
+ - Each source below keeps its newest version that passed every check; its newer batch waits until it passes them:
3508
+ - borealis-dataverse-workbooks-cc-by: changed pointer group file data/borealis-dataverse-workbooks-cc-by/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 2 key(s) of the old version are in no data file of source borealis-dataverse-workbooks-cc-by after the commit (row 4 from batch 2026-10-a-spreadsheets-with-answers, row 5 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3509
+ - borealis-dataverse-workbooks-cc0: changed pointer group file data/borealis-dataverse-workbooks-cc0/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 1 key(s) of the old version are in no data file of source borealis-dataverse-workbooks-cc0 after the commit (row 7 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3510
+ - cfpb-complaints: refused
3511
+ - data-gov-uk-workbooks: changed pointer group file data/data-gov-uk-workbooks/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 4 key(s) of the old version are in no data file of source data-gov-uk-workbooks after the commit (row 5 from batch 2026-10-a-spreadsheets-with-answers, row 6 from batch 2026-10-a-spreadsheets-with-answers, row 12 from batch 2026-10-a-spreadsheets-with-answers, 1 more); the removal record holds none of them, and none fails a publish check now
3512
+ - data-gov-uk-workbooks-noassertion: changed pointer group file data/data-gov-uk-workbooks-noassertion/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 1 key(s) of the old version are in no data file of source data-gov-uk-workbooks-noassertion after the commit (row 6 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3513
+ - dataversenl-workbooks-cc-by: changed pointer group file data/dataversenl-workbooks-cc-by/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 1 key(s) of the old version are in no data file of source dataversenl-workbooks-cc-by after the commit (row 17 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3514
+ - dataversenl-workbooks-cc0: changed pointer group file data/dataversenl-workbooks-cc0/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 3 key(s) of the old version are in no data file of source dataversenl-workbooks-cc0 after the commit (row 7 from batch 2026-10-a-spreadsheets-with-answers, row 13 from batch 2026-10-a-spreadsheets-with-answers, row 16 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3515
+ - dvn-fvhlgg-repro: changed pointer group file data/dvn-fvhlgg-repro/2026-10-c-repros/part-00000.parquet: 3 key(s) of the old version are in no data file of source dvn-fvhlgg-repro after the commit (row 5 from batch 2026-10-c-repros, row 7 from batch 2026-10-c-repros, row 16 from batch 2026-10-c-repros); the removal record holds none of them, and none fails a publish check now
3516
+ - eu-institution-workbooks-cc-by: changed pointer group file data/eu-institution-workbooks-cc-by/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 3 key(s) of the old version are in no data file of source eu-institution-workbooks-cc-by after the commit (row 2 from batch 2026-10-a-spreadsheets-with-answers, row 4 from batch 2026-10-a-spreadsheets-with-answers, row 5 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3517
+ - hdx-workbooks-cc0: changed pointer group file data/hdx-workbooks-cc0/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 2 key(s) of the old version are in no data file of source hdx-workbooks-cc0 after the commit (row 4 from batch 2026-10-a-spreadsheets-with-answers, row 5 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3518
+ - hdx-workbooks-odbl: changed pointer group file data/hdx-workbooks-odbl/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 1 key(s) of the old version are in no data file of source hdx-workbooks-odbl after the commit (row 4 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
3519
+ - hf-training-records: refused
3520
+ - npm: refused
3521
+ - scotland-government-workbooks: changed pointer group file data/scotland-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 2 key(s) of the old version are in no data file of source scotland-government-workbooks after the commit (row 3 from batch 2026-10-a-spreadsheets-with-answers-2, row 6 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3522
+ - stackexchange-api-posts: refused
3523
+ - uk-government-workbooks: changed pointer group file data/uk-government-workbooks/2026-10-a-spreadsheets-with-answers-2/part-00000.parquet: 1 key(s) of the old version are in no data file of source uk-government-workbooks after the commit (row 22 from batch 2026-10-a-spreadsheets-with-answers-2); the removal record holds none of them, and none fails a publish check now
3524
+ - victoria-government-workbooks: changed pointer group file data/victoria-government-workbooks/2026-10-a-spreadsheets-with-answers/part-00000.parquet: 2 key(s) of the old version are in no data file of source victoria-government-workbooks after the commit (row 6 from batch 2026-10-a-spreadsheets-with-answers, row 7 from batch 2026-10-a-spreadsheets-with-answers); the removal record holds none of them, and none fails a publish check now
README.md CHANGED
@@ -35,18 +35,18 @@ this dataset is the foundation's public index.
35
  English | [简体中文](README.zh.md) · [Data page](https://huggingface.co/spaces/Battam/training-data-foundation) · [Usage video](https://huggingface.co/spaces/Battam/training-data-foundation#usage-video)
36
 
37
  <picture class="not-prose">
38
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-narrow-light.svg">
39
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-wide-light.svg" alt="What each training stage can use in this release. Pre-training (provisional): 30.6 TB deduplicated size, 149,118,671 rows, 29.9 TB allowed for training. Mid-training (provisional): 3.86 TB deduplicated size, 34,084,498 rows, 1.42 TB allowed for training. Post-training: 1,763 material sets, from 1,820 upstream sources, 1,411 allowed for training.">
40
  </picture>
41
  <picture class="not-prose">
42
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-narrow-dark.svg">
43
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-wide-dark.svg" alt="What each training stage can use in this release. Pre-training (provisional): 30.6 TB deduplicated size, 149,118,671 rows, 29.9 TB allowed for training. Mid-training (provisional): 3.86 TB deduplicated size, 34,084,498 rows, 1.42 TB allowed for training. Post-training: 1,763 material sets, from 1,820 upstream sources, 1,411 allowed for training.">
44
  </picture>
45
 
46
  ## Get data in one command
47
 
48
  ```sh
49
- uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/tools/fetch.py --form spreadsheets --sample 3 --out ./data
50
  ```
51
 
52
  This fetches 3 units of every spreadsheet material set into `./data`, one folder per unit, and checks each file against the size and hashes its row records.
@@ -56,7 +56,7 @@ the same command takes the same units of this release.
56
  <details>
57
  <summary>What the command takes, and in which order</summary>
58
 
59
- The command and the queries below are pinned to the release tag `release-2026-11-12.9`.
60
  [uv](https://docs.astral.sh/uv/) reads the script's dependencies, at the versions published before the day of this release.
61
 
62
  The command returns units: 3 units from each material set whose form is `spreadsheets`, each in a folder of its own named after its `unit_id`.
@@ -84,7 +84,7 @@ Choose material sets from the catalog with [DuckDB](https://duckdb.org/):
84
 
85
  ```sql
86
  SELECT source, collection, title, units, files, top_types, license.tier AS license_tier, needs_auth
87
- FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet'
88
  WHERE kind = 'material_set' AND needs_auth = 'no' AND license.tier IN ('permissive', 'share-alike')
89
  ORDER BY collection, units DESC;
90
  ```
@@ -95,7 +95,7 @@ with `kind = 'corpus'` it returns corpora instead, counted in rows (`files`).
95
  Choose files in one corpus:
96
 
97
  ```sql
98
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'if-archive'));
99
  SELECT uri, size_bytes, lower(hex(sha256)) AS sha256
100
  FROM read_parquet(getvariable('files'))
101
  WHERE source = 'if-archive' AND ext = 'pdf' AND size_bytes BETWEEN 100000 AND 5000000
@@ -111,12 +111,12 @@ A data file may hold the rows of several sources, so the query keeps this one's
111
  ## What's inside
112
 
113
  <picture class="not-prose">
114
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-narrow-light.svg">
115
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-wide-light.svg" alt="The training stages by top-level domain, largest first. Pre-training (provisional), deduplicated size, by domain: Data science 29.8 TB; Science 531 GB; Machine learning 82.8 GB; Media 48.9 GB; System administration 35.7 GB; Hardware 29.6 GB; Many domains 29.6 GB; Security 4.18 GB; Unclassified 1.43 GB; Misc 376 MB; Software engineering 176 MB; Operations 4.08 KB. Mid-training (provisional), deduplicated size, by domain: System administration 1.27 TB; Many domains 1.00 TB; Machine learning 503 GB; Science 461 GB; Data science 429 GB; Hardware 177 GB; Security 13.9 GB; Misc 3.45 GB; Software engineering 2.66 GB; Unclassified 2.21 GB; Media 701 MB; Operations 13.8 MB. Post-training, material sets, by domain: Data science 348; Software engineering 286; Media 273; Operations 206; Machine learning 182; Science 174; Security 103; Hardware 87; System administration 71; Misc 33.">
116
  </picture>
117
  <picture class="not-prose">
118
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-narrow-dark.svg">
119
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-wide-dark.svg" alt="The training stages by top-level domain, largest first. Pre-training (provisional), deduplicated size, by domain: Data science 29.8 TB; Science 531 GB; Machine learning 82.8 GB; Media 48.9 GB; System administration 35.7 GB; Hardware 29.6 GB; Many domains 29.6 GB; Security 4.18 GB; Unclassified 1.43 GB; Misc 376 MB; Software engineering 176 MB; Operations 4.08 KB. Mid-training (provisional), deduplicated size, by domain: System administration 1.27 TB; Many domains 1.00 TB; Machine learning 503 GB; Science 461 GB; Data science 429 GB; Hardware 177 GB; Security 13.9 GB; Misc 3.45 GB; Software engineering 2.66 GB; Unclassified 2.21 GB; Media 701 MB; Operations 13.8 MB. Post-training, material sets, by domain: Data science 348; Software engineering 286; Media 273; Operations 206; Machine learning 182; Science 174; Security 103; Hardware 87; System administration 71; Misc 33.">
120
  </picture>
121
 
122
  <details>
@@ -133,7 +133,7 @@ A data file may hold the rows of several sources, so the query keeps this one's
133
  | Hardware | 29.6 GB | 388,557 | 177 GB | 4,945,550 | 87 |
134
  | Security | 4.18 GB | 80,666 | 13.9 GB | 117,641 | 103 |
135
  | Media | 48.9 GB | 16,365 | 701 MB | 34,670 | 273 |
136
- | Misc | 376 MB | 2,403 | 3.45 GB | 102,989 | 33 |
137
  | Many domains | 29.6 GB | 8,262,018 | 1.00 TB | 16,181,781 | 0 |
138
  | Unclassified | 1.43 GB | 336,597 | 2.21 GB | 401,574 | 0 |
139
 
@@ -148,12 +148,12 @@ Every material set and every corpus, with its counts, size, license tier and log
148
  ## How it is made
149
 
150
  <picture class="not-prose">
151
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-narrow-light.svg">
152
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-wide-light.svg" alt="How the index is made, in four steps. 1, collect: collectors read whole public upstreams into corpora, one row per file or record, the material of pre-training and mid-training. 2, admit: works the search found upstream and groups of corpus rows become material sets, the material of post-training. 3, check: licenses, personal data, overlap with benchmarks and sizes are checked before release. 4, release: one commit with its own tag, release-2026-11-12.9.">
153
  </picture>
154
  <picture class="not-prose">
155
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-narrow-dark.svg">
156
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-wide-dark.svg" alt="How the index is made, in four steps. 1, collect: collectors read whole public upstreams into corpora, one row per file or record, the material of pre-training and mid-training. 2, admit: works the search found upstream and groups of corpus rows become material sets, the material of post-training. 3, check: licenses, personal data, overlap with benchmarks and sizes are checked before release. 4, release: one commit with its own tag, release-2026-11-12.9.">
157
  </picture>
158
 
159
  Collectors read whole public upstreams into corpora, the material of pre-training and mid-training, and admission turns works found upstream and chosen corpus rows into material sets, the material of post-training.
@@ -181,7 +181,7 @@ The full text, with the license details, personal information read from sample f
181
  @misc{training_data_foundation,
182
  title = {training-data-foundation: an index of public materials for building agent environments},
183
  howpublished = {https://huggingface.co/datasets/Battam/training-data-foundation},
184
- note = {Release release-2026-11-12.9},
185
  year = {2026}
186
  }
187
  ```
 
35
  English | [简体中文](README.zh.md) · [Data page](https://huggingface.co/spaces/Battam/training-data-foundation) · [Usage video](https://huggingface.co/spaces/Battam/training-data-foundation#usage-video)
36
 
37
  <picture class="not-prose">
38
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/stages-narrow-light.svg">
39
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/stages-wide-light.svg" alt="What each training stage can use in this release. Pre-training (provisional): 30.6 TB deduplicated size, 149,118,671 rows, 29.9 TB allowed for training. Mid-training (provisional): 3.86 TB deduplicated size, 34,084,498 rows, 1.42 TB allowed for training. Post-training: 1,764 material sets, from 1,821 upstream sources, 1,412 allowed for training.">
40
  </picture>
41
  <picture class="not-prose">
42
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/stages-narrow-dark.svg">
43
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/stages-wide-dark.svg" alt="What each training stage can use in this release. Pre-training (provisional): 30.6 TB deduplicated size, 149,118,671 rows, 29.9 TB allowed for training. Mid-training (provisional): 3.86 TB deduplicated size, 34,084,498 rows, 1.42 TB allowed for training. Post-training: 1,764 material sets, from 1,821 upstream sources, 1,412 allowed for training.">
44
  </picture>
45
 
46
  ## Get data in one command
47
 
48
  ```sh
49
+ uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/tools/fetch.py --form spreadsheets --sample 3 --out ./data
50
  ```
51
 
52
  This fetches 3 units of every spreadsheet material set into `./data`, one folder per unit, and checks each file against the size and hashes its row records.
 
56
  <details>
57
  <summary>What the command takes, and in which order</summary>
58
 
59
+ The command and the queries below are pinned to the release tag `release-2026-11-12.10`.
60
  [uv](https://docs.astral.sh/uv/) reads the script's dependencies, at the versions published before the day of this release.
61
 
62
  The command returns units: 3 units from each material set whose form is `spreadsheets`, each in a folder of its own named after its `unit_id`.
 
84
 
85
  ```sql
86
  SELECT source, collection, title, units, files, top_types, license.tier AS license_tier, needs_auth
87
+ FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet'
88
  WHERE kind = 'material_set' AND needs_auth = 'no' AND license.tier IN ('permissive', 'share-alike')
89
  ORDER BY collection, units DESC;
90
  ```
 
95
  Choose files in one corpus:
96
 
97
  ```sql
98
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'if-archive'));
99
  SELECT uri, size_bytes, lower(hex(sha256)) AS sha256
100
  FROM read_parquet(getvariable('files'))
101
  WHERE source = 'if-archive' AND ext = 'pdf' AND size_bytes BETWEEN 100000 AND 5000000
 
111
  ## What's inside
112
 
113
  <picture class="not-prose">
114
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/inside-narrow-light.svg">
115
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/inside-wide-light.svg" alt="The training stages by top-level domain, largest first. Pre-training (provisional), deduplicated size, by domain: Data science 29.8 TB; Science 531 GB; Machine learning 82.8 GB; Media 48.9 GB; System administration 35.7 GB; Hardware 29.6 GB; Many domains 29.6 GB; Security 4.18 GB; Unclassified 1.43 GB; Misc 376 MB; Software engineering 176 MB; Operations 4.08 KB. Mid-training (provisional), deduplicated size, by domain: System administration 1.27 TB; Many domains 1.00 TB; Machine learning 503 GB; Science 461 GB; Data science 429 GB; Hardware 177 GB; Security 13.9 GB; Misc 3.45 GB; Software engineering 2.66 GB; Unclassified 2.21 GB; Media 701 MB; Operations 13.8 MB. Post-training, material sets, by domain: Data science 348; Software engineering 286; Media 273; Operations 206; Machine learning 182; Science 174; Security 103; Hardware 87; System administration 71; Misc 34.">
116
  </picture>
117
  <picture class="not-prose">
118
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/inside-narrow-dark.svg">
119
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/inside-wide-dark.svg" alt="The training stages by top-level domain, largest first. Pre-training (provisional), deduplicated size, by domain: Data science 29.8 TB; Science 531 GB; Machine learning 82.8 GB; Media 48.9 GB; System administration 35.7 GB; Hardware 29.6 GB; Many domains 29.6 GB; Security 4.18 GB; Unclassified 1.43 GB; Misc 376 MB; Software engineering 176 MB; Operations 4.08 KB. Mid-training (provisional), deduplicated size, by domain: System administration 1.27 TB; Many domains 1.00 TB; Machine learning 503 GB; Science 461 GB; Data science 429 GB; Hardware 177 GB; Security 13.9 GB; Misc 3.45 GB; Software engineering 2.66 GB; Unclassified 2.21 GB; Media 701 MB; Operations 13.8 MB. Post-training, material sets, by domain: Data science 348; Software engineering 286; Media 273; Operations 206; Machine learning 182; Science 174; Security 103; Hardware 87; System administration 71; Misc 34.">
120
  </picture>
121
 
122
  <details>
 
133
  | Hardware | 29.6 GB | 388,557 | 177 GB | 4,945,550 | 87 |
134
  | Security | 4.18 GB | 80,666 | 13.9 GB | 117,641 | 103 |
135
  | Media | 48.9 GB | 16,365 | 701 MB | 34,670 | 273 |
136
+ | Misc | 376 MB | 2,403 | 3.45 GB | 102,989 | 34 |
137
  | Many domains | 29.6 GB | 8,262,018 | 1.00 TB | 16,181,781 | 0 |
138
  | Unclassified | 1.43 GB | 336,597 | 2.21 GB | 401,574 | 0 |
139
 
 
148
  ## How it is made
149
 
150
  <picture class="not-prose">
151
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/steps-narrow-light.svg">
152
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/steps-wide-light.svg" alt="How the index is made, in four steps. 1, collect: collectors read whole public upstreams into corpora, one row per file or record, the material of pre-training and mid-training. 2, admit: works the search found upstream and groups of corpus rows become material sets, the material of post-training. 3, check: licenses, personal data, overlap with benchmarks and sizes are checked before release. 4, release: one commit with its own tag, release-2026-11-12.10.">
153
  </picture>
154
  <picture class="not-prose">
155
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/steps-narrow-dark.svg">
156
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/steps-wide-dark.svg" alt="How the index is made, in four steps. 1, collect: collectors read whole public upstreams into corpora, one row per file or record, the material of pre-training and mid-training. 2, admit: works the search found upstream and groups of corpus rows become material sets, the material of post-training. 3, check: licenses, personal data, overlap with benchmarks and sizes are checked before release. 4, release: one commit with its own tag, release-2026-11-12.10.">
157
  </picture>
158
 
159
  Collectors read whole public upstreams into corpora, the material of pre-training and mid-training, and admission turns works found upstream and chosen corpus rows into material sets, the material of post-training.
 
181
  @misc{training_data_foundation,
182
  title = {training-data-foundation: an index of public materials for building agent environments},
183
  howpublished = {https://huggingface.co/datasets/Battam/training-data-foundation},
184
+ note = {Release release-2026-11-12.10},
185
  year = {2026}
186
  }
187
  ```
README.zh.md CHANGED
@@ -22,18 +22,18 @@ tags:
22
  [English](README.md) | 简体中文 · [数据页](https://huggingface.co/spaces/Battam/training-data-foundation) · [使用示例视频](https://huggingface.co/spaces/Battam/training-data-foundation#usage-video)
23
 
24
  <picture class="not-prose">
25
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-narrow-light.zh.svg">
26
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-wide-light.zh.svg" alt="本次发布中每个训练阶段能用的数据。预训练(暂定):去重后大小 30.6 TB,149,118,671 行,其中 29.9 TB 允许用于训练。中训练(暂定):去重后大小 3.86 TB,34,084,498 行,其中 1.42 TB 允许用于训练。后训练:材料集 1,763,来自 1,820 个上游来源,其中 1,411 允许用于训练。">
27
  </picture>
28
  <picture class="not-prose">
29
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-narrow-dark.zh.svg">
30
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/stages-wide-dark.zh.svg" alt="本次发布中每个训练阶段能用的数据。预训练(暂定):去重后大小 30.6 TB,149,118,671 行,其中 29.9 TB 允许用于训练。中训练(暂定):去重后大小 3.86 TB,34,084,498 行,其中 1.42 TB 允许用于训练。后训练:材料集 1,763,来自 1,820 个上游来源,其中 1,411 允许用于训练。">
31
  </picture>
32
 
33
  ## 一条命令取数据
34
 
35
  ```sh
36
- uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/tools/fetch.py --form spreadsheets --sample 3 --out ./data
37
  ```
38
 
39
  这条命令从每个电子表格材料集各取 3 个单元放进 `./data`,每个单元一个文件夹,并按每个文件所在行记录的大小和哈希值校验它。
@@ -43,7 +43,7 @@ uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/r
43
  <details>
44
  <summary>命令取什么、按什么顺序取</summary>
45
 
46
- 这条命令和下面的查询都固定在发布标签 `release-2026-11-12.9` 上。
47
  [uv](https://docs.astral.sh/uv/) 会读取脚本声明的依赖,取这个发布当天之前发布的版本。
48
 
49
  命令返回单元:形式为 `spreadsheets` 的每个材料集各 3 个单元,每个单元放在以它的 `unit_id` 命名的单独文件夹里。
@@ -71,7 +71,7 @@ uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/r
71
 
72
  ```sql
73
  SELECT source, collection, title, units, files, top_types, license.tier AS license_tier, needs_auth
74
- FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet'
75
  WHERE kind = 'material_set' AND needs_auth = 'no' AND license.tier IN ('permissive', 'share-alike')
76
  ORDER BY collection, units DESC;
77
  ```
@@ -82,7 +82,7 @@ ORDER BY collection, units DESC;
82
  在一个语料库里挑选文件:
83
 
84
  ```sql
85
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'if-archive'));
86
  SELECT uri, size_bytes, lower(hex(sha256)) AS sha256
87
  FROM read_parquet(getvariable('files'))
88
  WHERE source = 'if-archive' AND ext = 'pdf' AND size_bytes BETWEEN 100000 AND 5000000
@@ -98,12 +98,12 @@ ORDER BY size_bytes LIMIT 20;
98
  ## 内容概览
99
 
100
  <picture class="not-prose">
101
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-narrow-light.zh.svg">
102
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-wide-light.zh.svg" alt="按顶层领域分的各训练阶段,由大到小。预训练(暂定),去重后大小,按领域:数据科学 29.8 TB;科学 531 GB;机器学习 82.8 GB;媒体 48.9 GB;系统管理 35.7 GB;硬件 29.6 GB;多领域 29.6 GB;安全 4.18 GB;未分类 1.43 GB;杂项 376 MB;软件工程 176 MB;运营 4.08 KB。中训练(暂定),去重后大小,按领域:系统管理 1.27 TB;多领域 1.00 TB;机器学习 503 GB;科学 461 GB;数据科学 429 GB;硬件 177 GB;安全 13.9 GB;杂项 3.45 GB;软件工程 2.66 GB;未分类 2.21 GB;媒体 701 MB;运营 13.8 MB。后训练,材料集,按领域:数据科学 348;软件工程 286;媒体 273;运营 206;机器学习 182;科学 174;安全 103;硬件 87;系统管理 71;杂项 33。">
103
  </picture>
104
  <picture class="not-prose">
105
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-narrow-dark.zh.svg">
106
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/inside-wide-dark.zh.svg" alt="按顶层领域分的各训练阶段,由大到小。预训练(暂定),去重后大小,按领域:数据科学 29.8 TB;科学 531 GB;机器学习 82.8 GB;媒体 48.9 GB;系统管理 35.7 GB;硬件 29.6 GB;多领域 29.6 GB;安全 4.18 GB;未分类 1.43 GB;杂项 376 MB;软件工程 176 MB;运营 4.08 KB。中训练(暂定),去重后大小,按领域:系统管理 1.27 TB;多领域 1.00 TB;机器学习 503 GB;科学 461 GB;数据科学 429 GB;硬件 177 GB;安全 13.9 GB;杂项 3.45 GB;软件工程 2.66 GB;未分类 2.21 GB;媒体 701 MB;运营 13.8 MB。后训练,材料集,按领域:数据科学 348;软件工程 286;媒体 273;运营 206;机器学习 182;科学 174;安全 103;硬件 87;系统管理 71;杂项 33。">
107
  </picture>
108
 
109
  <details>
@@ -120,7 +120,7 @@ ORDER BY size_bytes LIMIT 20;
120
  | 硬件 | 29.6 GB | 388,557 | 177 GB | 4,945,550 | 87 |
121
  | 安全 | 4.18 GB | 80,666 | 13.9 GB | 117,641 | 103 |
122
  | 媒体 | 48.9 GB | 16,365 | 701 MB | 34,670 | 273 |
123
- | 杂项 | 376 MB | 2,403 | 3.45 GB | 102,989 | 33 |
124
  | 多领域 | 29.6 GB | 8,262,018 | 1.00 TB | 16,181,781 | 0 |
125
  | 未分类 | 1.43 GB | 336,597 | 2.21 GB | 401,574 | 0 |
126
 
@@ -135,12 +135,12 @@ ORDER BY size_bytes LIMIT 20;
135
  ## 怎样做成
136
 
137
  <picture class="not-prose">
138
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-narrow-light.zh.svg">
139
- <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-wide-light.zh.svg" alt="索引分四步做成。1,采集:采集器整批读取完整的公开上游,每个文件或每条记录记一行,成为语料库,供预训练和中训练使用。2,收录:检索在上游找到的作品和成组的语料库行,收录为材料集,供后训练使用。3,检查:发布前核对许可证、个人信息、与基准测试的重叠和大小。4,发布:一个带有自己标签的提交,release-2026-11-12.9。">
140
  </picture>
141
  <picture class="not-prose">
142
- <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-narrow-dark.zh.svg">
143
- <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.9/docs/figures/steps-wide-dark.zh.svg" alt="索引分四步做成。1,采集:采集器整批读取完整的公开上游,每个文件或每条记录记一行,成为语料库,供预训练和中训练使用。2,收录:检索在上游找到的作品和成组的语料库行,收录为材料集,供后训练使用。3,检查:发布前核对许可证、个人信息、与基准测试的重叠和大小。4,发布:一个带有自己标签的提交,release-2026-11-12.9。">
144
  </picture>
145
 
146
  采集器整批读取完整的公开上游,记成语料库,供预训练和中训练使用,收录则把检索在上游找到的作品和选出的语料库行做成材料集,供后训练使用。
@@ -168,7 +168,7 @@ ORDER BY size_bytes LIMIT 20;
168
  @misc{training_data_foundation,
169
  title = {training-data-foundation: an index of public materials for building agent environments},
170
  howpublished = {https://huggingface.co/datasets/Battam/training-data-foundation},
171
- note = {Release release-2026-11-12.9},
172
  year = {2026}
173
  }
174
  ```
 
22
  [English](README.md) | 简体中文 · [数据页](https://huggingface.co/spaces/Battam/training-data-foundation) · [使用示例视频](https://huggingface.co/spaces/Battam/training-data-foundation#usage-video)
23
 
24
  <picture class="not-prose">
25
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/stages-narrow-light.zh.svg">
26
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/stages-wide-light.zh.svg" alt="本次发布中每个训练阶段能用的数据。预训练(暂定):去重后大小 30.6 TB,149,118,671 行,其中 29.9 TB 允许用于训练。中训练(暂定):去重后大小 3.86 TB,34,084,498 行,其中 1.42 TB 允许用于训练。后训练:材料集 1,764,来自 1,821 个上游来源,其中 1,412 允许用于训练。">
27
  </picture>
28
  <picture class="not-prose">
29
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/stages-narrow-dark.zh.svg">
30
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/stages-wide-dark.zh.svg" alt="本次发布中每个训练阶段能用的数据。预训练(暂定):去重后大小 30.6 TB,149,118,671 行,其中 29.9 TB 允许用于训练。中训练(暂定):去重后大小 3.86 TB,34,084,498 行,其中 1.42 TB 允许用于训练。后训练:材料集 1,764,来自 1,821 个上游来源,其中 1,412 允许用于训练。">
31
  </picture>
32
 
33
  ## 一条命令取数据
34
 
35
  ```sh
36
+ uv run https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/tools/fetch.py --form spreadsheets --sample 3 --out ./data
37
  ```
38
 
39
  这条命令从每个电子表格材料集各取 3 个单元放进 `./data`,每个单元一个文件夹,并按每个文件所在行记录的大小和哈希值校验它。
 
43
  <details>
44
  <summary>命令取什么、按什么顺序取</summary>
45
 
46
+ 这条命令和下面的查询都固定在发布标签 `release-2026-11-12.10` 上。
47
  [uv](https://docs.astral.sh/uv/) 会读取脚本声明的依赖,取这个发布当天之前发布的版本。
48
 
49
  命令返回单元:形式为 `spreadsheets` 的每个材料集各 3 个单元,每个单元放在以它的 `unit_id` 命名的单独文件夹里。
 
71
 
72
  ```sql
73
  SELECT source, collection, title, units, files, top_types, license.tier AS license_tier, needs_auth
74
+ FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet'
75
  WHERE kind = 'material_set' AND needs_auth = 'no' AND license.tier IN ('permissive', 'share-alike')
76
  ORDER BY collection, units DESC;
77
  ```
 
82
  在一个语料库里挑选文件:
83
 
84
  ```sql
85
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'if-archive'));
86
  SELECT uri, size_bytes, lower(hex(sha256)) AS sha256
87
  FROM read_parquet(getvariable('files'))
88
  WHERE source = 'if-archive' AND ext = 'pdf' AND size_bytes BETWEEN 100000 AND 5000000
 
98
  ## 内容概览
99
 
100
  <picture class="not-prose">
101
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/inside-narrow-light.zh.svg">
102
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/inside-wide-light.zh.svg" alt="按顶层领域分的各训练阶段,由大到小。预训练(暂定),去重后大小,按领域:数据科学 29.8 TB;科学 531 GB;机器学习 82.8 GB;媒体 48.9 GB;系统管理 35.7 GB;硬件 29.6 GB;多领域 29.6 GB;安全 4.18 GB;未分类 1.43 GB;杂项 376 MB;软件工程 176 MB;运营 4.08 KB。中训练(暂定),去重后大小,按领域:系统管理 1.27 TB;多领域 1.00 TB;机器学习 503 GB;科学 461 GB;数据科学 429 GB;硬件 177 GB;安全 13.9 GB;杂项 3.45 GB;软件工程 2.66 GB;未分类 2.21 GB;媒体 701 MB;运营 13.8 MB。后训练,材料集,按领域:数据科学 348;软件工程 286;媒体 273;运营 206;机器学习 182;科学 174;安全 103;硬件 87;系统管理 71;杂项 34。">
103
  </picture>
104
  <picture class="not-prose">
105
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/inside-narrow-dark.zh.svg">
106
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/inside-wide-dark.zh.svg" alt="按顶层领域分的各训练阶段,由大到小。预训练(暂定),去重后大小,按领域:数据科学 29.8 TB;科学 531 GB;机器学习 82.8 GB;媒体 48.9 GB;系统管理 35.7 GB;硬件 29.6 GB;多领域 29.6 GB;安全 4.18 GB;未分类 1.43 GB;杂项 376 MB;软件工程 176 MB;运营 4.08 KB。中训练(暂定),去重后大小,按领域:系统管理 1.27 TB;多领域 1.00 TB;机器学习 503 GB;科学 461 GB;数据科学 429 GB;硬件 177 GB;安全 13.9 GB;杂项 3.45 GB;软件工程 2.66 GB;未分类 2.21 GB;媒体 701 MB;运营 13.8 MB。后训练,材料集,按领域:数据科学 348;软件工程 286;媒体 273;运营 206;机器学习 182;科学 174;安全 103;硬件 87;系统管理 71;杂项 34。">
107
  </picture>
108
 
109
  <details>
 
120
  | 硬件 | 29.6 GB | 388,557 | 177 GB | 4,945,550 | 87 |
121
  | 安全 | 4.18 GB | 80,666 | 13.9 GB | 117,641 | 103 |
122
  | 媒体 | 48.9 GB | 16,365 | 701 MB | 34,670 | 273 |
123
+ | 杂项 | 376 MB | 2,403 | 3.45 GB | 102,989 | 34 |
124
  | 多领域 | 29.6 GB | 8,262,018 | 1.00 TB | 16,181,781 | 0 |
125
  | 未分类 | 1.43 GB | 336,597 | 2.21 GB | 401,574 | 0 |
126
 
 
135
  ## 怎样做成
136
 
137
  <picture class="not-prose">
138
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/steps-narrow-light.zh.svg">
139
+ <img class="not-prose dark:hidden" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/steps-wide-light.zh.svg" alt="索引分四步做成。1,采集:采集器整批读取完整的公开上游,每个文件或每条记录记一行,成为语料库,供预训练和中训练使用。2,收录:检索在上游找到的作品和成组的语料库行,收录为材料集,供后训练使用。3,检查:发布前核对许可证、个人信息、与基准测试的重叠和大小。4,发布:一个带有自己标签的提交,release-2026-11-12.10。">
140
  </picture>
141
  <picture class="not-prose">
142
+ <source media="(max-width: 1023px)" srcset="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/steps-narrow-dark.zh.svg">
143
+ <img class="not-prose hidden dark:block" src="https://huggingface.co/datasets/Battam/training-data-foundation/resolve/release-2026-11-12.10/docs/figures/steps-wide-dark.zh.svg" alt="索引分四步做成。1,采集:采集器整批读取完整的公开上游,每个文件或每条记录记一行,成为语料库,供预训练和中训练使用。2,收录:检索在上游找到的作品和成组的语料库行,收录为材料集,供后训练使用。3,检查:发布前核对许可证、个人信息、与基准测试的重叠和大小。4,发布:一个带有自己标签的提交,release-2026-11-12.10。">
144
  </picture>
145
 
146
  采集器整批读取完整的公开上游,记成语料库,供预训练和中训练使用,收录则把检索在上游找到的作品和选出的语料库行做成材料集,供后训练使用。
 
168
  @misc{training_data_foundation,
169
  title = {training-data-foundation: an index of public materials for building agent environments},
170
  howpublished = {https://huggingface.co/datasets/Battam/training-data-foundation},
171
+ note = {Release release-2026-11-12.10},
172
  year = {2026}
173
  }
174
  ```
campaign_files/2026-10-a-spreadsheets-with-answers-2/world-bank-workbooks-odbl/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** 1 of 2 files are self-contained; the others are used together with other upstream files.
10
 
11
- **Answers:** Each file with an answer has one answer file, at answer/<unit path>.json (the unit path plus .json). Reference result kind: key_formula_values.
12
-
13
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
14
 
15
  **Licence (SPDX):** ODbL-1.0
 
8
 
9
  **Self-contained:** 1 of 2 files are self-contained; the others are used together with other upstream files.
10
 
 
 
11
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx)
12
 
13
  **Licence (SPDX):** ODbL-1.0
campaign_files/2026-10-a-spreadsheets-with-answers-2/world-bank-workbooks/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** 31 of 42 files are self-contained; the others are used together with other upstream files.
10
 
11
- **Answers:** Each file with an answer has one answer file, at answer/<unit path>.json (the unit path plus .json). Reference result kind: key_formula_values.
12
-
13
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx); LibreOffice headless conversion (soffice --headless --convert-to xlsx), then openpyxl; xlrd to read values
14
 
15
  **Licence (SPDX):** CC-BY-4.0
 
8
 
9
  **Self-contained:** 31 of 42 files are self-contained; the others are used together with other upstream files.
10
 
 
 
11
  **Tools to open and check:** openpyxl (formulas and cached values); LibreOffice headless recalculation (soffice --headless --convert-to xlsx); LibreOffice headless conversion (soffice --headless --convert-to xlsx), then openpyxl; xlrd to read values
12
 
13
  **Licence (SPDX):** CC-BY-4.0
campaign_files/2026-10-g-solver-instances-2/xbandrade-sokoban/LICENSE ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ Licence: MIT
2
+
3
+ - MIT: https://github.com/xbandrade/sokoban-solver-generator/blob/69c7b5c8f8a8190a7e9c8bdaaa1771d13dfa9274/LICENSE (publisher: xbandrade)
campaign_files/2026-10-g-solver-instances-2/xbandrade-sokoban/README.md ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # xbandrade-sokoban
2
+
3
+ **What this is:** 1 unit (12 files) published by xbandrade. Collection theme: Seeded puzzle generators (Simon Tatham puzzle collection class, with generation command, seed and solution).
4
+
5
+ **What one unit is:** Each unit is a few named upstream files, delivered together, kept at its upstream path. Format: Text grid per level plus solution move string (JSONL index), plus a generation record.
6
+
7
+ **Where it comes from:** Upstream: xbandrade/sokoban-solver-generator. Kind: generated data.
8
+
9
+ **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
+
11
+ **Tools to open and check:** Check: Replay the submitted moves from the level with the upstream can_move (src/utils.py line 169) and require is_solved (src/utils.py lines 52-53: no box off a goal is left) at the end; solutions are not unique, so no exact match. Wrong-answer controls (truncated move string, an illegal move) must fail.
12
+
13
+ **Licence (SPDX):** MIT
14
+
15
+ **Provenance:** https://github.com/xbandrade/sokoban-solver-generator
16
+
17
+ Units:
18
+
19
+ - `/` (12 files): xbandrade/sokoban-solver-generator: seeded reverse-construction Sokoban level generator with BFS and A* solvers.
20
+ Format: text grid per level plus solution move string (JSONL index), plus a generation record
21
+ Check: Replay the submitted moves from the level with the upstream can_move (src/utils.py line 169) and require is_solved (src/utils.py lines 52-53: no box off a goal is left) at the end; solutions are not unique, so no exact match. Wrong-answer controls (truncated move string, an illegal move) must fail.
22
+ Companions (not included): src/utils.py; src/bfs.py; src/astar.py; src/game.py; src/player.py; src/box.py; src/floor.py; sokoban.py; requirements.txt; LICENSE; README.md
23
+ Dependencies: pygame==2.2.0 (requirements.txt, UTF-16); numpy; matplotlib==3.7.1, pygame-widgets==1.1.1, pytest==7.2.2 (requirements.txt; not needed by generate); xbandrade/sokoban-solver-generator at git commit 69c7b5c8 with its img/ assets (generator loads them); not packed
campaign_files/2026-10-g-solver-instances-2/xcsp3-checking-instances/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each file is self-contained. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Complete solution set (enumerate all solutions); number of solutions (model counting); published optimum per instance (COP). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Exact: the agent's set of solutions (as XCSP3 instantiations) must equal the README set for that file, compared as sets of value tuples; a single-solution task accepts any member of the set; Exact integer match of the agent's solution count with readme.txt. Exact: the agent's assignment must satisfy every constraint of the instance (checked by an XCSP3 solution checker) and its objective value must equal the README optimum for that instance; for 2023 counting cases the reported number of solutions must equal the xls count.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,21 +19,15 @@ Files:
21
  - `g043-01/checkingInstancesXCSP25.zip`: XCSP3 2025 checking instances: 4 small CSP instances that exercise rarely used XCSP3 features, with every solution listed.
22
  Format: zip with TestingInstances2025/README and 4 XCSP3 XML instances (TestSetConditions, TestOrderedVarLengths, TestNonBinaryOperators, TestMixedTerms); a case is one XML instance; its answer is the block of <instantiation> lines for that file in README
23
  Origin: hand-written test instances published by the XCSP3 organiser for solver authors to check parser and constraint support before the 2025 competition
24
- Reference result: complete solution set (enumerate all solutions)
25
- Reference source: likewise lines 11, 18 and 27 for the other three files (4, 6 and 6 solutions). Published by the competition organiser.
26
  Check: Exact: the agent's set of solutions (as XCSP3 instantiations) must equal the README set for that file, compared as sets of value tuples; a single-solution task accepts any member of the set
27
  Dependencies: not packed
28
  - `g043-02/checkingInstancesXCSP24.zip`: XCSP3 2024 checking instances: 4 CSP instances (binary puzzles, lex with constant limit, undefined variables) with published solution counts.
29
  Format: zip with checkingInstancesXCSP24/readme.txt and 4 XCSP3 XML instances (BinaryPuzzle-4, BinaryPuzzle-6, TestLexCst, TestUndefined); a case is one XML instance; its answer is the solution count in readme.txt
30
  Origin: organiser-written checking instances for the XCSP3 2024 competition
31
- Reference result: number of solutions (model counting)
32
- Reference source: checkingInstancesXCSP24/readme.txt lines 2, 3, 5 and 7 (72, 4140, 293 and 6 solutions). Published by the organiser.
33
  Check: Exact integer match of the agent's solution count with readme.txt.
34
  Dependencies: not packed
35
  - `g184-01/checkingInstancesXCSP23.zip`: XCSP3 2023 checking instances: COP and CSP instances for solver authors, 14 of them (2022/Inst01-14) with the optimum stated in README and the 2023 tests with optimum or solution count in tests2023.xls.
36
  Format: zip of 312 entries: 2017/, 2018/, 2022/ (Inst01-14.xml.lzma with README.txt), 2023/ (14 XML files with README.txt and tests2023.xls), all/csp (92) and all/cop (47) .xml.lzma; a case is one instance; LZMA members unpack with Python's lzma module
37
  Origin: instances published by the XCSP3 organiser for solver authors to check parser and constraint support before the 2022 and 2023 competitions
38
- Reference result: published optimum per instance (COP)
39
- Reference source: 2022/README.txt inside the zip gives the optimum of each of Inst01 to Inst14 (lines 15 to 114, e.g. line 15 "** Inst01 (optimum 38)"); top README.txt line 1 says the 2023 directory has "known results (number of solutions or optimum)", kept in 2023/tests2023.xls (binary XLS; strings show e.g. "ObjVarCoeffs-int Opt=56", "Obj-varCoeffs-var Opt=40"). Published by the competition organiser.
40
  Check: Exact: the agent's assignment must satisfy every constraint of the instance (checked by an XCSP3 solution checker) and its objective value must equal the README optimum for that instance; for 2023 counting cases the reported number of solutions must equal the xls count
41
  Dependencies: not packed; Python lzma module of the standard library to unpack .xml.lzma members
 
8
 
9
  **Self-contained:** Each file is self-contained. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Exact: the agent's set of solutions (as XCSP3 instantiations) must equal the README set for that file, compared as sets of value tuples; a single-solution task accepts any member of the set; Exact integer match of the agent's solution count with readme.txt. Exact: the agent's assignment must satisfy every constraint of the instance (checked by an XCSP3 solution checker) and its objective value must equal the README optimum for that instance; for 2023 counting cases the reported number of solutions must equal the xls count.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `g043-01/checkingInstancesXCSP25.zip`: XCSP3 2025 checking instances: 4 small CSP instances that exercise rarely used XCSP3 features, with every solution listed.
20
  Format: zip with TestingInstances2025/README and 4 XCSP3 XML instances (TestSetConditions, TestOrderedVarLengths, TestNonBinaryOperators, TestMixedTerms); a case is one XML instance; its answer is the block of <instantiation> lines for that file in README
21
  Origin: hand-written test instances published by the XCSP3 organiser for solver authors to check parser and constraint support before the 2025 competition
 
 
22
  Check: Exact: the agent's set of solutions (as XCSP3 instantiations) must equal the README set for that file, compared as sets of value tuples; a single-solution task accepts any member of the set
23
  Dependencies: not packed
24
  - `g043-02/checkingInstancesXCSP24.zip`: XCSP3 2024 checking instances: 4 CSP instances (binary puzzles, lex with constant limit, undefined variables) with published solution counts.
25
  Format: zip with checkingInstancesXCSP24/readme.txt and 4 XCSP3 XML instances (BinaryPuzzle-4, BinaryPuzzle-6, TestLexCst, TestUndefined); a case is one XML instance; its answer is the solution count in readme.txt
26
  Origin: organiser-written checking instances for the XCSP3 2024 competition
 
 
27
  Check: Exact integer match of the agent's solution count with readme.txt.
28
  Dependencies: not packed
29
  - `g184-01/checkingInstancesXCSP23.zip`: XCSP3 2023 checking instances: COP and CSP instances for solver authors, 14 of them (2022/Inst01-14) with the optimum stated in README and the 2023 tests with optimum or solution count in tests2023.xls.
30
  Format: zip of 312 entries: 2017/, 2018/, 2022/ (Inst01-14.xml.lzma with README.txt), 2023/ (14 XML files with README.txt and tests2023.xls), all/csp (92) and all/cop (47) .xml.lzma; a case is one instance; LZMA members unpack with Python's lzma module
31
  Origin: instances published by the XCSP3 organiser for solver authors to check parser and constraint support before the 2022 and 2023 competitions
 
 
32
  Check: Exact: the agent's assignment must satisfy every constraint of the instance (checked by an XCSP3 solution checker) and its objective value must equal the README optimum for that instance; for 2023 counting cases the reported number of solutions must equal the xls count
33
  Dependencies: not packed; Python lzma module of the standard library to unpack .xml.lzma members
campaign_files/2026-10-g-solver-instances-2/xhstt-archive/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each file is self-contained. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Proven optimum (objective equals lower bound or zero). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Program: run the pinned XHSTT evaluator on the agent's solution for instance ZA-LW-09; require infeasibility cost 0 and objective cost at most 0; Program: run the pinned XHSTT evaluator on the agent's solution for instance BrazilInstance1_XHSTT-v2014; require infeasibility cost 0 and objective cost at most 41.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,14 +19,10 @@ Files:
21
  - `SouthAfrica/Lewitt2009/SouthAfricaLewitt2009.xml`: XHSTT Lewitt2009 primary school timetable (South Africa): 148 times, 19 teachers, 2 rooms, 16 classes, 185 events; optimum cost (0, 0).
22
  Format: XHSTT XML archive: one Instance (lines 1 to 4893) and SolutionGroups after it; when posing, strip the SolutionGroups element so only the instance is given
23
  Origin: XHSTT benchmark archive (Post et al.)
24
- Reference result: proven optimum (objective equals lower bound or zero)
25
- Reference source: dataset page SouthAfrica/Lewitt2009/: solution table row 'VAGO2012 0 0 Christos Gogos et al. April 14, 2012' (feasibility value 0, objective value 0); the same solution group sits in the XML from line 7478
26
  Check: Program: run the pinned XHSTT evaluator on the agent's solution for instance ZA-LW-09; require infeasibility cost 0 and objective cost at most 0
27
  Dependencies: XHSTT evaluator HSEval by J. Kingston
28
  - `Brazil/Instance1/BrazilInstance1.xml`: XHSTT BrazilInstance1 high school timetable: 25 times, 8 teachers, 3 classes, 21 events; optimum objective 41.
29
  Format: XHSTT XML archive: one Instance (lines 1 to 1115) and SolutionGroups after it; when posing, strip the SolutionGroups element so only the instance is given
30
  Origin: XHSTT benchmark archive (Post et al.)
31
- Reference result: proven optimum (objective equals lower bound or zero)
32
- Reference source: dataset page Brazil/Instance1/: solution table row 'LectioIntegerProgramming 41 41 Matias Sørensen April 4, 2015' (objective value 41, lower bound 41); solutions in the XML from line 1116
33
  Check: Program: run the pinned XHSTT evaluator on the agent's solution for instance BrazilInstance1_XHSTT-v2014; require infeasibility cost 0 and objective cost at most 41
34
  Dependencies: XHSTT evaluator HSEval by J. Kingston
 
8
 
9
  **Self-contained:** Each file is self-contained. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Program: run the pinned XHSTT evaluator on the agent's solution for instance ZA-LW-09; require infeasibility cost 0 and objective cost at most 0; Program: run the pinned XHSTT evaluator on the agent's solution for instance BrazilInstance1_XHSTT-v2014; require infeasibility cost 0 and objective cost at most 41.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `SouthAfrica/Lewitt2009/SouthAfricaLewitt2009.xml`: XHSTT Lewitt2009 primary school timetable (South Africa): 148 times, 19 teachers, 2 rooms, 16 classes, 185 events; optimum cost (0, 0).
20
  Format: XHSTT XML archive: one Instance (lines 1 to 4893) and SolutionGroups after it; when posing, strip the SolutionGroups element so only the instance is given
21
  Origin: XHSTT benchmark archive (Post et al.)
 
 
22
  Check: Program: run the pinned XHSTT evaluator on the agent's solution for instance ZA-LW-09; require infeasibility cost 0 and objective cost at most 0
23
  Dependencies: XHSTT evaluator HSEval by J. Kingston
24
  - `Brazil/Instance1/BrazilInstance1.xml`: XHSTT BrazilInstance1 high school timetable: 25 times, 8 teachers, 3 classes, 21 events; optimum objective 41.
25
  Format: XHSTT XML archive: one Instance (lines 1 to 1115) and SolutionGroups after it; when posing, strip the SolutionGroups element so only the instance is given
26
  Origin: XHSTT benchmark archive (Post et al.)
 
 
27
  Check: Program: run the pinned XHSTT evaluator on the agent's solution for instance BrazilInstance1_XHSTT-v2014; require infeasibility cost 0 and objective cost at most 41
28
  Dependencies: XHSTT evaluator HSEval by J. Kingston
campaign_files/2026-10-g-solver-instances-2/xhstt-high-school-timetabling/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Best known objective (soft cost at zero infeasibility). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Program: evaluate the agent's XHSTT solution with an XHSTT evaluator (e.g. Kingston's HSEval, linked from the HSTT index page); require infeasibility cost 0 and objective cost at most 101.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,7 +19,5 @@ Files:
21
  - `datasets/USA/Westside2009/USAWestside2009.xml`: XHSTT high school timetabling instance USA Westside 2009 with recorded solutions; best soft cost 101 (Skolaris, 2017) at zero infeasibility.
22
  Format: XHSTT XML archive: one Instance (lines 1 to 26760) and SolutionGroups (lines 26761 to 164710, five groups incl. Skolaris_2017-09-13 at line 136582); when posing, strip the SolutionGroups element so only the instance is shown; one case
23
  Origin: real US high school timetable (Westside 2009) anonymised and published in the XHSTT archive
24
- Reference result: best known objective (soft cost at zero infeasibility)
25
- Reference source: westside.html (single-line HTML, line 1) results table row "Skolaris 101 101 Martin Klemsa September 13, 2017"; the solution itself is SolutionGroup Skolaris_2017-09-13 at line 136582 of the XML
26
  Check: Program: evaluate the agent's XHSTT solution with an XHSTT evaluator (e.g. Kingston's HSEval, linked from the HSTT index page); require infeasibility cost 0 and objective cost at most 101
27
  Dependencies: XHSTT evaluator HSEval by J. Kingston
 
8
 
9
  **Self-contained:** The file is self-contained. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Program: evaluate the agent's XHSTT solution with an XHSTT evaluator (e.g. Kingston's HSEval, linked from the HSTT index page); require infeasibility cost 0 and objective cost at most 101.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `datasets/USA/Westside2009/USAWestside2009.xml`: XHSTT high school timetabling instance USA Westside 2009 with recorded solutions; best soft cost 101 (Skolaris, 2017) at zero infeasibility.
20
  Format: XHSTT XML archive: one Instance (lines 1 to 26760) and SolutionGroups (lines 26761 to 164710, five groups incl. Skolaris_2017-09-13 at line 136582); when posing, strip the SolutionGroups element so only the instance is shown; one case
21
  Origin: real US high school timetable (Westside 2009) anonymised and published in the XHSTT archive
 
 
22
  Check: Program: evaluate the agent's XHSTT solution with an XHSTT evaluator (e.g. Kingston's HSEval, linked from the HSTT index page); require infeasibility cost 0 and objective cost at most 101
23
  Dependencies: XHSTT evaluator HSEval by J. Kingston
campaign_files/2026-10-gt-browser-tests-2/x-wr-timezone-single-events/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_output_files. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Python x-wr-timezone (to_standard) on an icalendar Calendar, or its command line `x-wr-timezone in.ics out.ics`; a script rewrites the UTC times into the calendar's X-WR-TIMEZONE zone and writes the new .ics
14
 
15
  **Licence (SPDX):** LGPL-3.0-only
@@ -21,8 +19,6 @@ Files:
21
  - `test/calendars/single-events-DTSTART-DTEND.in.ics`: Convert UTC event times into the calendar's X-WR-TIMEZONE zone.
22
  Format: iCalendar input and expected output (.ics pair); pytest files
23
  Origin: x-wr-timezone conversion test pair (tool_fixture)
24
- Reference result: expected_output_files
25
- Reference source: test/calendars/single-events-DTSTART-DTEND.out.ics is the expected converted calendar (test_convert_examples.py line 21); lines 39-45 pin DTSTART;TZID=America/New_York:20211222T120000, DTEND 20211222T130000, DTSTART 20211222T210000 and DTEND 20211222T220000
26
  Check: Final file state: the agent writes the converted calendar to a file; parsed with icalendar it must compare equal to the parsed .out.ics, and its DTSTART and DTEND lines must carry TZID=America/New_York with the four expected local times. No screenshot
27
  Tools: Python x-wr-timezone (to_standard) on an icalendar Calendar, or its command line `x-wr-timezone in.ics out.ics`; a script rewrites the UTC times into the calendar's X-WR-TIMEZONE zone and writes the new .ics
28
  Companions (not included): test/test_convert_examples.py; test/conftest.py; test/calendars/single-events-DTSTART-DTEND.out.ics
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Python x-wr-timezone (to_standard) on an icalendar Calendar, or its command line `x-wr-timezone in.ics out.ics`; a script rewrites the UTC times into the calendar's X-WR-TIMEZONE zone and writes the new .ics
12
 
13
  **Licence (SPDX):** LGPL-3.0-only
 
19
  - `test/calendars/single-events-DTSTART-DTEND.in.ics`: Convert UTC event times into the calendar's X-WR-TIMEZONE zone.
20
  Format: iCalendar input and expected output (.ics pair); pytest files
21
  Origin: x-wr-timezone conversion test pair (tool_fixture)
 
 
22
  Check: Final file state: the agent writes the converted calendar to a file; parsed with icalendar it must compare equal to the parsed .out.ics, and its DTSTART and DTEND lines must carry TZID=America/New_York with the four expected local times. No screenshot
23
  Tools: Python x-wr-timezone (to_standard) on an icalendar Calendar, or its command line `x-wr-timezone in.ics out.ics`; a script rewrites the UTC times into the calendar's X-WR-TIMEZONE zone and writes the new .ics
24
  Companions (not included): test/test_convert_examples.py; test/conftest.py; test/calendars/single-events-DTSTART-DTEND.out.ics
campaign_files/2026-10-m-issue-trackers-2/webrtc-gerrit-changes/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the change's own review timeline (status, submit time and submitter, patch sets, votes, reviewers added or removed, review comments). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** curl the Gerrit REST API (GET https://webrtc-review.googlesource.com/changes/498801/detail), strip the first line ")]}'" and read the JSON with jq or Python json; the git-review CLI and Gerrit's ssh 'gerrit query --format=JSON --all-approvals --comments' reach the same data where ssh access is open
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,7 +19,5 @@ Files:
21
  - `changes/498801/detail.json`: Src change 498801: Add audio_level, absolute_capture_timestamp and csrcs overrides to audio injection path.
22
  Format: Gerrit REST API ChangeInfo JSON from the /detail endpoint, preceded by the anti-XSSI line ")]}'": status, created, updated, submitted, submitter, owner, labels with every vote, messages (author, real_author, date, tag, _revision_number, message text) and reviewer_updates (updated, updated_by, reviewer, state REVIEWER/CC/REMOVED). Patch-set uploads are messages tagged autogenerated:gerrit:newPatchSet; votes appear in message text as 'Patch Set N: Label+V' and in labels.*.all. The endpoint returns all messages and reviewer updates of the change, so the record is complete.
23
  Origin: Live Gerrit REST interface of a closed (merged) code review change, collected as a NOASSERTION pointer row under spec general rule 4: the detail JSON was fetched twice apart and was byte-identical. Unit file kept only outside the repository for checking and then deleted.
24
- Reference result: facts computed from the change's own review timeline (status, submit time and submitter, patch sets, votes, reviewers added or removed, review comments)
25
- Reference source: The pinned response: changes/498801/detail.json.
26
  Check: Exact match against values read from the pinned JSON: status, submitted and submitter; reviewer_updates entries (reviewer, state, updated_by, updated); messages tagged autogenerated:gerrit:newPatchSet (patch-set count, _revision_number); votes from labels.*.all (value, _account_id) and from message first lines 'Patch Set N: Label+V'. A human review comment is a message whose author is not a bot or service account, whose tag does not start with 'autogenerated:', and whose text has content beyond the 'Patch Set N: votes' first line or names inline comments '(N comments)'. Bot and service accounts excluded from the comment count: webrtc-scoped@luci-project-accounts.iam.gserviceaccount.com, accounts tagged SERVICE_USER, and any account whose name, username or email matches the pattern (bot|zuul|jenkins|ci|luci|gerrit code review|service user|automation|buildbot|lint|sonar|tryjob|cq|autosubmit|copybara|tricium|gerritforge|verifier|checker|scheduler|maloo|janitor|autotest|gerrit). Deterministic, no model judgement. Before answers are issued, the extraction program must pass on 2 hand-checked units of this source (spec general rule 5); hand-checking is pending for the delivery step.
27
  Tools: curl the Gerrit REST API (GET https://webrtc-review.googlesource.com/changes/498801/detail), strip the first line ")]}'" and read the JSON with jq or Python json; the git-review CLI and Gerrit's ssh 'gerrit query --format=JSON --all-approvals --comments' reach the same data where ssh access is open
 
8
 
9
  **Self-contained:** The file is self-contained.
10
 
 
 
11
  **Tools to open and check:** curl the Gerrit REST API (GET https://webrtc-review.googlesource.com/changes/498801/detail), strip the first line ")]}'" and read the JSON with jq or Python json; the git-review CLI and Gerrit's ssh 'gerrit query --format=JSON --all-approvals --comments' reach the same data where ssh access is open
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `changes/498801/detail.json`: Src change 498801: Add audio_level, absolute_capture_timestamp and csrcs overrides to audio injection path.
20
  Format: Gerrit REST API ChangeInfo JSON from the /detail endpoint, preceded by the anti-XSSI line ")]}'": status, created, updated, submitted, submitter, owner, labels with every vote, messages (author, real_author, date, tag, _revision_number, message text) and reviewer_updates (updated, updated_by, reviewer, state REVIEWER/CC/REMOVED). Patch-set uploads are messages tagged autogenerated:gerrit:newPatchSet; votes appear in message text as 'Patch Set N: Label+V' and in labels.*.all. The endpoint returns all messages and reviewer updates of the change, so the record is complete.
21
  Origin: Live Gerrit REST interface of a closed (merged) code review change, collected as a NOASSERTION pointer row under spec general rule 4: the detail JSON was fetched twice apart and was byte-identical. Unit file kept only outside the repository for checking and then deleted.
 
 
22
  Check: Exact match against values read from the pinned JSON: status, submitted and submitter; reviewer_updates entries (reviewer, state, updated_by, updated); messages tagged autogenerated:gerrit:newPatchSet (patch-set count, _revision_number); votes from labels.*.all (value, _account_id) and from message first lines 'Patch Set N: Label+V'. A human review comment is a message whose author is not a bot or service account, whose tag does not start with 'autogenerated:', and whose text has content beyond the 'Patch Set N: votes' first line or names inline comments '(N comments)'. Bot and service accounts excluded from the comment count: webrtc-scoped@luci-project-accounts.iam.gserviceaccount.com, accounts tagged SERVICE_USER, and any account whose name, username or email matches the pattern (bot|zuul|jenkins|ci|luci|gerrit code review|service user|automation|buildbot|lint|sonar|tryjob|cq|autosubmit|copybara|tricium|gerritforge|verifier|checker|scheduler|maloo|janitor|autotest|gerrit). Deterministic, no model judgement. Before answers are issued, the extraction program must pass on 2 hand-checked units of this source (spec general rule 5); hand-checking is pending for the delivery step.
23
  Tools: curl the Gerrit REST API (GET https://webrtc-review.googlesource.com/changes/498801/detail), strip the first line ")]}'" and read the JSON with jq or Python json; the git-review CLI and Gerrit's ssh 'gerrit query --format=JSON --all-approvals --comments' reach the same data where ssh access is open
campaign_files/2026-10-m-issue-trackers-2/westcat-gtfs/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the service calendar itself (active services on a date, removed or added dates, day counts). Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -22,8 +20,6 @@ Units:
22
  Read out of the archive: http://data.trilliumtransit.com/gtfs/westcat-ca-us/westcat-ca-us.zip
23
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt, feed_info.txt)
24
  Origin: Live static GTFS feed published by WestCAT (Western Contra Costa Transit Authority, California), feed produced and hosted by Trillium Solutions at http://data.trilliumtransit.com/gtfs/westcat-ca-us/westcat-ca-us.zip. calendar.txt and its companions were read by member out of the zip with HTTP Range requests (spec general rule 4: by member); kept as a snapshot pinned by sha256.
25
- Reference result: facts computed from the service calendar itself (active services on a date, removed or added dates, day counts)
26
- Reference source: calendar.txt weekday flags with start_date and end_date, calendar_dates.txt exception rows, agency_timezone of agency.txt, all from the pinned members.
27
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
28
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
29
  Companions (not included): calendar_dates.txt; agency.txt; feed_info.txt
 
8
 
9
  **Self-contained:** The unit is self-contained. 1 of 1 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
12
 
13
  **Licence (SPDX):** NOASSERTION
 
20
  Read out of the archive: http://data.trilliumtransit.com/gtfs/westcat-ca-us/westcat-ca-us.zip
21
  Format: text/csv (GTFS Schedule calendar.txt; companions calendar_dates.txt, agency.txt, feed_info.txt)
22
  Origin: Live static GTFS feed published by WestCAT (Western Contra Costa Transit Authority, California), feed produced and hosted by Trillium Solutions at http://data.trilliumtransit.com/gtfs/westcat-ca-us/westcat-ca-us.zip. calendar.txt and its companions were read by member out of the zip with HTTP Range requests (spec general rule 4: by member); kept as a snapshot pinned by sha256.
 
 
23
  Check: deterministic, no model judgement. Hand-checking 2 samples per source is pending for the delivery step (spec general rule 5).
24
  Tools: Read the CSV files with Python csv or pandas (or sqlite3 .import) and expand service days per GTFS rules; gtfs-kit or partridge compute active service_ids per date headless; dates are local to agency_timezone (Python zoneinfo).
25
  Companions (not included): calendar_dates.txt; agency.txt; feed_info.txt
campaign_files/2026-10-m-issue-trackers-2/whamcloud-jira/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each file is self-contained.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the ticket's own timeline (status transitions, assignee changes, comments). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** curl the Jira REST API v2 (/rest/api/2/issue/KEY?expand=changelog) and read it with jq or Python json; the jira Python library or the Atlassian CLI also work.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,14 +19,10 @@ Files:
21
  - `rest/api/2/issue/LU-15000.json`: Whamcloud Jira LU-15000: MDS crash with ASSERTION(osp->opd_connects == 1) in osp_obd_connect (Resolved, Fixed, 2022-05-30).
22
  Format: Jira REST API v2 issue JSON with expand=changelog: fields (status, resolution, assignee, reporter, comment.comments[]) and changelog.histories[] (author, created, items[] with field, fromString, toString).
23
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
24
- Reference result: facts computed from the ticket's own timeline (status transitions, assignee changes, comments)
25
- Reference source: The pinned response(s): rest/api/2/issue/LU-15000.json.
26
  Check: Exact match against values read from the pinned JSON: changelog.histories[].items[] with field 'status' or 'assignee' (fromString, toString, created, author.displayName) and fields.comment.comments[] (author.displayName, created). Deterministic, no model judgement. Human comment count excludes bot accounts: Gerrit Updater. Timeline events = comments plus changelog histories. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
27
  Tools: curl the Jira REST API v2 (/rest/api/2/issue/KEY?expand=changelog) and read it with jq or Python json; the jira Python library or the Atlassian CLI also work.
28
  - `rest/api/2/issue/LU-16500.json`: Whamcloud Jira LU-16500: lfs migrate <file> preserves specific layout too much (Resolved, Fixed; reopened once).
29
  Format: Jira REST API v2 issue JSON with expand=changelog: fields (status, resolution, assignee, reporter, comment.comments[]) and changelog.histories[] (author, created, items[] with field, fromString, toString).
30
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
31
- Reference result: facts computed from the ticket's own timeline (status transitions, assignee changes, comments)
32
- Reference source: The pinned response(s): rest/api/2/issue/LU-16500.json.
33
  Check: Exact match against values read from the pinned JSON: changelog.histories[].items[] with field 'status' or 'assignee' (fromString, toString, created, author.displayName) and fields.comment.comments[] (author.displayName, created). Deterministic, no model judgement. Human comment count excludes bot accounts: Gerrit Updater. Timeline events = comments plus changelog histories. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
34
  Tools: curl the Jira REST API v2 (/rest/api/2/issue/KEY?expand=changelog) and read it with jq or Python json; the jira Python library or the Atlassian CLI also work.
 
8
 
9
  **Self-contained:** Each file is self-contained.
10
 
 
 
11
  **Tools to open and check:** curl the Jira REST API v2 (/rest/api/2/issue/KEY?expand=changelog) and read it with jq or Python json; the jira Python library or the Atlassian CLI also work.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `rest/api/2/issue/LU-15000.json`: Whamcloud Jira LU-15000: MDS crash with ASSERTION(osp->opd_connects == 1) in osp_obd_connect (Resolved, Fixed, 2022-05-30).
20
  Format: Jira REST API v2 issue JSON with expand=changelog: fields (status, resolution, assignee, reporter, comment.comments[]) and changelog.histories[] (author, created, items[] with field, fromString, toString).
21
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
 
 
22
  Check: Exact match against values read from the pinned JSON: changelog.histories[].items[] with field 'status' or 'assignee' (fromString, toString, created, author.displayName) and fields.comment.comments[] (author.displayName, created). Deterministic, no model judgement. Human comment count excludes bot accounts: Gerrit Updater. Timeline events = comments plus changelog histories. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
23
  Tools: curl the Jira REST API v2 (/rest/api/2/issue/KEY?expand=changelog) and read it with jq or Python json; the jira Python library or the Atlassian CLI also work.
24
  - `rest/api/2/issue/LU-16500.json`: Whamcloud Jira LU-16500: lfs migrate <file> preserves specific layout too much (Resolved, Fixed; reopened once).
25
  Format: Jira REST API v2 issue JSON with expand=changelog: fields (status, resolution, assignee, reporter, comment.comments[]) and changelog.histories[] (author, created, items[] with field, fromString, toString).
26
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
 
 
27
  Check: Exact match against values read from the pinned JSON: changelog.histories[].items[] with field 'status' or 'assignee' (fromString, toString, created, author.displayName) and fields.comment.comments[] (author.displayName, created). Deterministic, no model judgement. Human comment count excludes bot accounts: Gerrit Updater. Timeline events = comments plus changelog histories. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
28
  Tools: curl the Jira REST API v2 (/rest/api/2/issue/KEY?expand=changelog) and read it with jq or Python json; the jira Python library or the Atlassian CLI also work.
campaign_files/2026-10-m-issue-trackers-2/whatwg/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the thread headers. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Read the archive as mbox with Python's mailbox and email modules (or mutt -f / notmuch on the decompressed file); rebuild the thread from Message-ID, In-Reply-To and References headers.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,7 +19,5 @@ Files:
21
  - `m329-05/2014-March.txt.gz`: [whatwg] addPath and CanvasPathMethods (11 messages, 2014-03-20 to 2014-03-21).
22
  Format: Pipermail month archive, mbox text gzip-compressed: full headers (From with obfuscated address and name, Date, Subject, Message-ID, In-Reply-To, References) and plain-text bodies. Cut the thread by root Message-ID through the In-Reply-To closure (last References entry when In-Reply-To is missing).
23
  Origin: Real list traffic in a closed past month of the list's archive. The unit is one thread, cut out of the month file by its root Message-ID; the month file is the pinned pointer and is delivered whole. The landing page notes a gap in the static archive in 2014-2017 (a complete archive is kept by the W3C); 2014-March.txt.gz is present with 222 messages and the thread is complete against 2014-April.txt.gz. The month file has Last-Modified Tue, 06 Oct 2026 (regenerated recently), so copy B was the deciding pin check; it matched.
24
- Reference result: facts computed from the thread headers
25
- Reference source: Message-ID, In-Reply-To, References, From and Date headers of the messages in the thread, read from the pinned month file.
26
  Check: Exact match against values computed from the pinned month file: parse it as mbox, drop repeated copies of a Message-ID already seen (first copy kept), take the root Message-ID, collect every message whose In-Reply-To (or last References entry when In-Reply-To is missing) chains back to it, then count, take dates in UTC (a Date header without zone is read as UTC) and compare sender names. Message-IDs are compared after removing angle brackets and double quotes. People are counted by person: From entries with the same address, or the same name after removing accents and letter case, are one person (the names in scale are the first spelling of each). The first-reply interval is the exact difference in whole seconds between the Date headers of the root and the earliest reply, converted to UTC; Date headers have one-second resolution, so no rounding is involved. Deterministic. The program was checked on 2 threads of this source by a second, independent parser (splits the raw text on mbox From lines and reads headers with regular expressions, no mailbox or email module): the chosen thread (also read message by message) and '[whatwg] Canvas Path.addPath SVGMatrix not optimal?' (10 messages, 4 people, root <CE08235E-C01E-4001-BA46-4CACA2E6D6FC@adobe.com>); message count, Message-ID set, last message time and first-reply seconds agreed on both.
27
  Tools: Read the archive as mbox with Python's mailbox and email modules (or mutt -f / notmuch on the decompressed file); rebuild the thread from Message-ID, In-Reply-To and References headers.
 
8
 
9
  **Self-contained:** The file is self-contained.
10
 
 
 
11
  **Tools to open and check:** Read the archive as mbox with Python's mailbox and email modules (or mutt -f / notmuch on the decompressed file); rebuild the thread from Message-ID, In-Reply-To and References headers.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `m329-05/2014-March.txt.gz`: [whatwg] addPath and CanvasPathMethods (11 messages, 2014-03-20 to 2014-03-21).
20
  Format: Pipermail month archive, mbox text gzip-compressed: full headers (From with obfuscated address and name, Date, Subject, Message-ID, In-Reply-To, References) and plain-text bodies. Cut the thread by root Message-ID through the In-Reply-To closure (last References entry when In-Reply-To is missing).
21
  Origin: Real list traffic in a closed past month of the list's archive. The unit is one thread, cut out of the month file by its root Message-ID; the month file is the pinned pointer and is delivered whole. The landing page notes a gap in the static archive in 2014-2017 (a complete archive is kept by the W3C); 2014-March.txt.gz is present with 222 messages and the thread is complete against 2014-April.txt.gz. The month file has Last-Modified Tue, 06 Oct 2026 (regenerated recently), so copy B was the deciding pin check; it matched.
 
 
22
  Check: Exact match against values computed from the pinned month file: parse it as mbox, drop repeated copies of a Message-ID already seen (first copy kept), take the root Message-ID, collect every message whose In-Reply-To (or last References entry when In-Reply-To is missing) chains back to it, then count, take dates in UTC (a Date header without zone is read as UTC) and compare sender names. Message-IDs are compared after removing angle brackets and double quotes. People are counted by person: From entries with the same address, or the same name after removing accents and letter case, are one person (the names in scale are the first spelling of each). The first-reply interval is the exact difference in whole seconds between the Date headers of the root and the earliest reply, converted to UTC; Date headers have one-second resolution, so no rounding is involved. Deterministic. The program was checked on 2 threads of this source by a second, independent parser (splits the raw text on mbox From lines and reads headers with regular expressions, no mailbox or email module): the chosen thread (also read message by message) and '[whatwg] Canvas Path.addPath SVGMatrix not optimal?' (10 messages, 4 people, root <CE08235E-C01E-4001-BA46-4CACA2E6D6FC@adobe.com>); message count, Message-ID set, last message time and first-reply seconds agreed on both.
23
  Tools: Read the archive as mbox with Python's mailbox and email modules (or mutt -f / notmuch on the decompressed file); rebuild the thread from Message-ID, In-Reply-To and References headers.
campaign_files/2026-10-m-issue-trackers-2/wikimedia-gerrit-changes/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the change's own review timeline (status, submit time and submitter, patch sets, votes, reviewers added or removed, review comments). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** curl the Gerrit REST API (GET https://gerrit.wikimedia.org/r/changes/{number}/detail), strip the first line ")]}'" and read the JSON with jq or Python json; the git-review CLI and Gerrit's ssh 'gerrit query --format=JSON --all-approvals --comments' reach the same data where ssh access is open
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,7 +19,5 @@ Files:
21
  - `changes/1144576/detail.json`: mediawiki/core change 1144576: block: Deprecate UserIdentity|string block target specifications (take 2).
22
  Format: Gerrit REST API ChangeInfo JSON from the /detail endpoint, preceded by the anti-XSSI line ")]}'": status, created, updated, submitted, submitter, owner, labels with every vote (DETAILED_LABELS), messages (author, real_author, date, tag, _revision_number, message text) and reviewer_updates (updated, updated_by, reviewer, state REVIEWER/CC/REMOVED). Patch-set uploads are messages tagged autogenerated:gerrit:newPatchSet; votes appear in message text as 'Patch Set N: Label+V' and in labels.*.all. The endpoint returns all messages and reviewer updates of the change, so the record is complete.
23
  Origin: Live Gerrit REST interface of a closed (merged) code review change, collected as a NOASSERTION pointer row under spec general rule 4: the detail JSON was fetched twice apart and was byte-identical. Unit file kept only outside the repository for checking and then deleted.
24
- Reference result: facts computed from the change's own review timeline (status, submit time and submitter, patch sets, votes, reviewers added or removed, review comments)
25
- Reference source: The pinned response: changes/1144576/detail.json.
26
  Check: Exact match against values read from the pinned JSON: status, submitted and submitter; reviewer_updates entries (reviewer, state, updated_by, updated); messages tagged autogenerated:gerrit:newPatchSet (patch-set count, _revision_number); votes from labels.*.all (value, _account_id) and from message first lines 'Patch Set N: Label+V'. A human review comment is a message whose author is not a bot or service account, whose tag does not start with 'autogenerated:', and whose text has content beyond the 'Patch Set N: votes' first line or names inline comments '(N comments)'. Bot and service accounts excluded from the comment count: jenkins-bot, sonarqubebot, accounts tagged SERVICE_USER, and any account whose name, username or email matches the pattern (bot|zuul|jenkins|ci|luci|gerrit code review|service user|automation|buildbot|lint|sonar|tryjob|cq|autosubmit|copybara|tricium|gerritforge|verifier|checker|scheduler). Deterministic, no model judgement. Before answers are issued, the extraction program must pass on 2 hand-checked units of this source (spec general rule 5); hand-checking is pending for the delivery step.
27
  Tools: curl the Gerrit REST API (GET https://gerrit.wikimedia.org/r/changes/{number}/detail), strip the first line ")]}'" and read the JSON with jq or Python json; the git-review CLI and Gerrit's ssh 'gerrit query --format=JSON --all-approvals --comments' reach the same data where ssh access is open
 
8
 
9
  **Self-contained:** The file is self-contained.
10
 
 
 
11
  **Tools to open and check:** curl the Gerrit REST API (GET https://gerrit.wikimedia.org/r/changes/{number}/detail), strip the first line ")]}'" and read the JSON with jq or Python json; the git-review CLI and Gerrit's ssh 'gerrit query --format=JSON --all-approvals --comments' reach the same data where ssh access is open
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `changes/1144576/detail.json`: mediawiki/core change 1144576: block: Deprecate UserIdentity|string block target specifications (take 2).
20
  Format: Gerrit REST API ChangeInfo JSON from the /detail endpoint, preceded by the anti-XSSI line ")]}'": status, created, updated, submitted, submitter, owner, labels with every vote (DETAILED_LABELS), messages (author, real_author, date, tag, _revision_number, message text) and reviewer_updates (updated, updated_by, reviewer, state REVIEWER/CC/REMOVED). Patch-set uploads are messages tagged autogenerated:gerrit:newPatchSet; votes appear in message text as 'Patch Set N: Label+V' and in labels.*.all. The endpoint returns all messages and reviewer updates of the change, so the record is complete.
21
  Origin: Live Gerrit REST interface of a closed (merged) code review change, collected as a NOASSERTION pointer row under spec general rule 4: the detail JSON was fetched twice apart and was byte-identical. Unit file kept only outside the repository for checking and then deleted.
 
 
22
  Check: Exact match against values read from the pinned JSON: status, submitted and submitter; reviewer_updates entries (reviewer, state, updated_by, updated); messages tagged autogenerated:gerrit:newPatchSet (patch-set count, _revision_number); votes from labels.*.all (value, _account_id) and from message first lines 'Patch Set N: Label+V'. A human review comment is a message whose author is not a bot or service account, whose tag does not start with 'autogenerated:', and whose text has content beyond the 'Patch Set N: votes' first line or names inline comments '(N comments)'. Bot and service accounts excluded from the comment count: jenkins-bot, sonarqubebot, accounts tagged SERVICE_USER, and any account whose name, username or email matches the pattern (bot|zuul|jenkins|ci|luci|gerrit code review|service user|automation|buildbot|lint|sonar|tryjob|cq|autosubmit|copybara|tricium|gerritforge|verifier|checker|scheduler). Deterministic, no model judgement. Before answers are issued, the extraction program must pass on 2 hand-checked units of this source (spec general rule 5); hand-checking is pending for the delivery step.
23
  Tools: curl the Gerrit REST API (GET https://gerrit.wikimedia.org/r/changes/{number}/detail), strip the first line ")]}'" and read the JSON with jq or Python json; the git-review CLI and Gerrit's ssh 'gerrit query --format=JSON --all-approvals --comments' reach the same data where ssh access is open
campaign_files/2026-10-m-issue-trackers-2/wildfiregames-trac/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each file is self-contained.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the ticket's own timeline (status transitions, assignee changes, comments). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** curl the Trac ticket RSS export (/ticket/N?format=rss) and parse it with xmllint, xmlstarlet or Python xml.etree; Trac also offers ?format=csv and ?format=tab for the current field values.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,21 +19,15 @@ Files:
21
  - `ticket/3000.rss`: 0 A.D. Trac #3000: [PATCH] GUI: hero icon area limited to one hero at a time (closed, fixed).
22
  Format: Trac ticket RSS 2.0 export: one <item> per ticket change (dc:creator, pubDate, title naming the changed fields, HTML description with the field changes and the comment text).
23
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
24
- Reference result: facts computed from the ticket's own timeline (status transitions, assignee changes, comments)
25
- Reference source: The pinned response(s): ticket/3000.rss.
26
  Check: Exact match against values read from the pinned RSS: each <item> is one timeline entry (dc:creator, pubDate, title naming changed fields, description with the field list 'status changed from X to Y' / 'owner set to Z' and the comment text after it). Deterministic, no model judgement. Human comments = items with comment text after the field list, excluding attachment-only entries, empty or placeholder text and commit-hook notices (comment text starting 'In <changeset>' or items without dc:creator); no bot accounts were seen on this tracker. Timeline events = RSS items. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
27
  Tools: curl the Trac ticket RSS export (/ticket/N?format=rss) and parse it with xmllint, xmlstarlet or Python xml.etree; Trac also offers ?format=csv and ?format=tab for the current field values.
28
  - `ticket/2000.rss`: 0 A.D. Trac #2000: OOS in a multiplayer game against Aegis (closed after a reopen).
29
  Format: Trac ticket RSS 2.0 export: one <item> per ticket change (dc:creator, pubDate, title naming the changed fields, HTML description with the field changes and the comment text).
30
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
31
- Reference result: facts computed from the ticket's own timeline (status transitions, assignee changes, comments)
32
- Reference source: The pinned response(s): ticket/2000.rss.
33
  Check: Exact match against values read from the pinned RSS: each <item> is one timeline entry (dc:creator, pubDate, title naming changed fields, description with the field list 'status changed from X to Y' / 'owner set to Z' and the comment text after it). Deterministic, no model judgement. Human comments = items with comment text after the field list, excluding attachment-only entries, empty or placeholder text and commit-hook notices (comment text starting 'In <changeset>' or items without dc:creator); no bot accounts were seen on this tracker. Timeline events = RSS items. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
34
  Tools: curl the Trac ticket RSS export (/ticket/N?format=rss) and parse it with xmllint, xmlstarlet or Python xml.etree; Trac also offers ?format=csv and ?format=tab for the current field values.
35
  - `ticket/3500.rss`: 0 A.D. Trac #3500: [PATCH] Compilation warnings on OS X 10.9.5 (closed, fixed).
36
  Format: Trac ticket RSS 2.0 export: one <item> per ticket change (dc:creator, pubDate, title naming the changed fields, HTML description with the field changes and the comment text).
37
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
38
- Reference result: facts computed from the ticket's own timeline (status transitions, assignee changes, comments)
39
- Reference source: The pinned response(s): ticket/3500.rss.
40
  Check: Exact match against values read from the pinned RSS: each <item> is one timeline entry (dc:creator, pubDate, title naming changed fields, description with the field list 'status changed from X to Y' / 'owner set to Z' and the comment text after it). Deterministic, no model judgement. Human comments = items with comment text after the field list, excluding attachment-only entries, empty or placeholder text and commit-hook notices (comment text starting 'In <changeset>' or items without dc:creator); no bot accounts were seen on this tracker. Timeline events = RSS items. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
41
  Tools: curl the Trac ticket RSS export (/ticket/N?format=rss) and parse it with xmllint, xmlstarlet or Python xml.etree; Trac also offers ?format=csv and ?format=tab for the current field values.
 
8
 
9
  **Self-contained:** Each file is self-contained.
10
 
 
 
11
  **Tools to open and check:** curl the Trac ticket RSS export (/ticket/N?format=rss) and parse it with xmllint, xmlstarlet or Python xml.etree; Trac also offers ?format=csv and ?format=tab for the current field values.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `ticket/3000.rss`: 0 A.D. Trac #3000: [PATCH] GUI: hero icon area limited to one hero at a time (closed, fixed).
20
  Format: Trac ticket RSS 2.0 export: one <item> per ticket change (dc:creator, pubDate, title naming the changed fields, HTML description with the field changes and the comment text).
21
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
 
 
22
  Check: Exact match against values read from the pinned RSS: each <item> is one timeline entry (dc:creator, pubDate, title naming changed fields, description with the field list 'status changed from X to Y' / 'owner set to Z' and the comment text after it). Deterministic, no model judgement. Human comments = items with comment text after the field list, excluding attachment-only entries, empty or placeholder text and commit-hook notices (comment text starting 'In <changeset>' or items without dc:creator); no bot accounts were seen on this tracker. Timeline events = RSS items. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
23
  Tools: curl the Trac ticket RSS export (/ticket/N?format=rss) and parse it with xmllint, xmlstarlet or Python xml.etree; Trac also offers ?format=csv and ?format=tab for the current field values.
24
  - `ticket/2000.rss`: 0 A.D. Trac #2000: OOS in a multiplayer game against Aegis (closed after a reopen).
25
  Format: Trac ticket RSS 2.0 export: one <item> per ticket change (dc:creator, pubDate, title naming the changed fields, HTML description with the field changes and the comment text).
26
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
 
 
27
  Check: Exact match against values read from the pinned RSS: each <item> is one timeline entry (dc:creator, pubDate, title naming changed fields, description with the field list 'status changed from X to Y' / 'owner set to Z' and the comment text after it). Deterministic, no model judgement. Human comments = items with comment text after the field list, excluding attachment-only entries, empty or placeholder text and commit-hook notices (comment text starting 'In <changeset>' or items without dc:creator); no bot accounts were seen on this tracker. Timeline events = RSS items. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
28
  Tools: curl the Trac ticket RSS export (/ticket/N?format=rss) and parse it with xmllint, xmlstarlet or Python xml.etree; Trac also offers ?format=csv and ?format=tab for the current field values.
29
  - `ticket/3500.rss`: 0 A.D. Trac #3500: [PATCH] Compilation warnings on OS X 10.9.5 (closed, fixed).
30
  Format: Trac ticket RSS 2.0 export: one <item> per ticket change (dc:creator, pubDate, title naming the changed fields, HTML description with the field changes and the comment text).
31
  Origin: Live tracker interface of a closed ticket, collected as a NOASSERTION pointer row under spec general rule 4: copies fetched apart were byte-identical. Unit files kept only outside the repository for checking and then deleted.
 
 
32
  Check: Exact match against values read from the pinned RSS: each <item> is one timeline entry (dc:creator, pubDate, title naming changed fields, description with the field list 'status changed from X to Y' / 'owner set to Z' and the comment text after it). Deterministic, no model judgement. Human comments = items with comment text after the field list, excluding attachment-only entries, empty or placeholder text and commit-hook notices (comment text starting 'In <changeset>' or items without dc:creator); no bot accounts were seen on this tracker. Timeline events = RSS items. Before answers are issued, the extraction program must pass on 2 hand-checked tickets of this tracker (spec general rule 5); hand-checking is pending for the delivery step.
33
  Tools: curl the Trac ticket RSS export (/ticket/N?format=rss) and parse it with xmllint, xmlstarlet or Python xml.etree; Trac also offers ?format=csv and ?format=tab for the current field values.
campaign_files/2026-10-m-issue-trackers-2/xapian-trac/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained.
10
 
11
- **Answers:** No answer files are included. Reference result: Facts computed from the ticket's own timeline (status transitions, assignee changes, comments). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** curl the Trac ticket RSS (https://trac.xapian.org/ticket/N?format=rss) or CSV (?format=csv) and parse it with Python xml.etree or csv.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,7 +19,5 @@ Files:
21
  - `ticket/300.rss`: Xapian Trac #300: Test failure : apitest, bigoaddvalue (closed, fixed, 2008-10-27).
22
  Format: Trac ticket RSS 2.0: one item per change (dc:creator, pubDate, title naming the changed fields, description holding field changes such as owner, status, resolution, milestone, plus the comment text).
23
  Origin: Live Trac ticket change feed, collected as a NOASSERTION pointer row under spec general rule 4: two copies fetched apart were byte-identical. Unit file kept only outside the repository for checking and then deleted.
24
- Reference result: facts computed from the ticket's own timeline (status transitions, assignee changes, comments)
25
- Reference source: The pinned feed: ticket/300.rss.
26
  Check: Exact match against values parsed from the pinned RSS items (dc:creator, pubDate, and the field changes in description: 'owner changed from X to Y', 'status A → B', 'resolution → R'). Deterministic, no model judgement. Human comment count excludes bot accounts: no bot account appears in this ticket (all item authors were checked against the regex bot|automation|jenkins|gerrit|system|migrat|importer; none matched). Extraction program to be hand-checked on 2 Trac tickets before answers are issued (general rule 5); pending.
27
  Tools: curl the Trac ticket RSS (https://trac.xapian.org/ticket/N?format=rss) or CSV (?format=csv) and parse it with Python xml.etree or csv.
 
8
 
9
  **Self-contained:** The file is self-contained.
10
 
 
 
11
  **Tools to open and check:** curl the Trac ticket RSS (https://trac.xapian.org/ticket/N?format=rss) or CSV (?format=csv) and parse it with Python xml.etree or csv.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `ticket/300.rss`: Xapian Trac #300: Test failure : apitest, bigoaddvalue (closed, fixed, 2008-10-27).
20
  Format: Trac ticket RSS 2.0: one item per change (dc:creator, pubDate, title naming the changed fields, description holding field changes such as owner, status, resolution, milestone, plus the comment text).
21
  Origin: Live Trac ticket change feed, collected as a NOASSERTION pointer row under spec general rule 4: two copies fetched apart were byte-identical. Unit file kept only outside the repository for checking and then deleted.
 
 
22
  Check: Exact match against values parsed from the pinned RSS items (dc:creator, pubDate, and the field changes in description: 'owner changed from X to Y', 'status A → B', 'resolution → R'). Deterministic, no model judgement. Human comment count excludes bot accounts: no bot account appears in this ticket (all item authors were checked against the regex bot|automation|jenkins|gerrit|system|migrat|importer; none matched). Extraction program to be hand-checked on 2 Trac tickets before answers are issued (general rule 5); pending.
23
  Tools: curl the Trac ticket RSS (https://trac.xapian.org/ticket/N?format=rss) or CSV (?format=csv) and parse it with Python xml.etree or csv.
campaign_files/2026-10-md-media-and-documents/when-in-rome-openscore-lieder/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each unit is self-contained. 36 of 36 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: roman_numeral_analysis_file. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
14
 
15
  **Licence (SPDX):** CC-BY-SA-4.0 AND CC0-1.0
@@ -21,16 +19,12 @@ Units:
21
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/01_Gute_Nacht/score.mxl`: Roman numeral analysis of Schubert, Winterreise D.911 No.1 'Gute Nacht' from the OpenScore score.
22
  Format: directory of 5 files: score.mxl (compressed MusicXML, 29,857 bytes), analysis.txt (RomanText, 1,509 bytes), plus automatic rntxt, dez and mscz
23
  Origin: human expert analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
24
- Reference result: roman_numeral_analysis_file
25
- Reference source: analysis.txt (RomanText), hand analysis by Mark Gotham (header: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students')
26
  Check: hide analysis.txt (and analysis_automatic.rntxt); the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key and Roman numeral (figure plus inversion) accuracy over the score's beats; report a key-only and a chord-only score. Expert analyses legitimately differ, so calibrate the band with a second annotator's corpus before setting a pass line
27
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
28
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez; lc5015378.mscz (MuseScore source of the same score)
29
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/05_Der_Lindenbaum/` (4 files): Roman numeral analysis of Schubert, Winterreise D.911 No.5 "Der Lindenbaum" from the OpenScore score.
30
  Format: directory of 5 files, 101,599 bytes: score.mxl (compressed MusicXML), analysis.txt (RomanText, with Form: lines), analysis_automatic.rntxt, analysis_dez_format.dez, lc5016466.mscz
31
  Origin: human expert analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
32
- Reference result: roman_numeral_analysis_file
33
- Reference source: analysis.txt (RomanText), hand analysis (header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.')
34
  Check: hide analysis.txt and analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band (for example the Winterreise Dataset annotations of the same song) is measured before any pass line is set; the Form: lines give a separate section-labelling task
35
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
36
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez; lc5016466.mscz (MuseScore source of the same score)
@@ -38,8 +32,6 @@ Units:
38
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/13_Die_Post/` (4 files): Roman numeral analysis of Schubert, Winterreise D.911 No.13 "Die Post" from the OpenScore score.
39
  Format: directory of 5 files, 89,732 bytes: score.mxl (compressed MusicXML), analysis.txt (RomanText, with Form: lines), analysis_automatic.rntxt, analysis_dez_format.dez, lc5007176.mscz
40
  Origin: human expert analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
41
- Reference result: roman_numeral_analysis_file
42
- Reference source: analysis.txt (RomanText), hand analysis (header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.')
43
  Check: hide analysis.txt and analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band (for example the Winterreise Dataset annotations of the same song) is measured before any pass line is set; the Form: lines give a separate section-labelling task
44
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
45
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez; lc5007176.mscz (MuseScore source of the same score)
@@ -47,8 +39,6 @@ Units:
47
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/20_Der_Wegweiser/` (4 files): Roman numeral analysis of Schubert, Winterreise D.911 No.20 "Der Wegweiser" from the OpenScore score.
48
  Format: directory of 5 files, 84,589 bytes: score.mxl (compressed MusicXML), analysis.txt (RomanText), analysis_automatic.rntxt, analysis_dez_format.dez, lc5009900.mscz
49
  Origin: human expert analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
50
- Reference result: roman_numeral_analysis_file
51
- Reference source: analysis.txt (RomanText), hand analysis (header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.')
52
  Check: hide analysis.txt and analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band (for example the Winterreise Dataset annotations of the same song) is measured before any pass line is set
53
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
54
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez; lc5009900.mscz (MuseScore source of the same score)
@@ -56,8 +46,6 @@ Units:
56
  - `Corpus/OpenScore-LiederCorpus/Coleridge-Taylor,_Samuel/6_Sorrow_Songs,_Op.57/1_Oh_what_comes_over_the_Sea/` (4 files): Roman numeral analysis of Samuel Coleridge-Taylor, 6 Sorrow Songs, Op.57 - No.1: Oh what comes over the Sea, from the OpenScore score.
57
  Format: directory of 5 files, 53,322 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6189652.mscz, score.mxl
58
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
59
- Reference result: roman_numeral_analysis_file
60
- Reference source: analysis.txt (RomanText), 21 measure lines; header line 3: 'Analyst: S. Reenan and M. Gotham'
61
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
62
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
63
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6189652.mscz (MuseScore source of the same score)
@@ -65,8 +53,6 @@ Units:
65
  - `Corpus/OpenScore-LiederCorpus/Burleigh,_Harry_Thacker/5_Songs_of_Laurence_Hope/5_Till_I_Wake/` (4 files): Roman numeral analysis of Harry Thacker Burleigh, 5 Songs of Laurence Hope - No.5: Till I Wake, from the OpenScore score.
66
  Format: directory of 5 files, 77,398 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6471199.mscz, score.mxl
67
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
68
- Reference result: roman_numeral_analysis_file
69
- Reference source: analysis.txt (RomanText), 84 measure lines; header line 3: 'Analyst: Mark Gotham'
70
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
71
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
72
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6471199.mscz (MuseScore source of the same score)
@@ -74,8 +60,6 @@ Units:
74
  - `Corpus/OpenScore-LiederCorpus/Chaminade,_Cécile/_/Berceuse/` (4 files): Roman numeral analysis of Cécile Chaminade, Berceuse, from the OpenScore score.
75
  Format: directory of 5 files, 69,112 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6264559.mscz, score.mxl
76
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
77
- Reference result: roman_numeral_analysis_file
78
- Reference source: analysis.txt (RomanText), 36 measure lines; header line 3: 'Analyst: Anonymous contributor via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/'; Proofreader: Mark Gotham
79
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
80
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
81
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6264559.mscz (MuseScore source of the same score)
@@ -83,8 +67,6 @@ Units:
83
  - `Corpus/OpenScore-LiederCorpus/Chausson,_Ernest/7_Mélodies,_Op.2/7_Le_Colibri/` (4 files): Roman numeral analysis of Ernest Chausson, 7 Mélodies, Op.2 - 7: Le Colibri, from the OpenScore score.
84
  Format: directory of 5 files, 67,039 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5077780.mscz, score.mxl
85
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
86
- Reference result: roman_numeral_analysis_file
87
- Reference source: analysis.txt (RomanText), 35 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
88
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
89
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
90
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5077780.mscz (MuseScore source of the same score)
@@ -92,8 +74,6 @@ Units:
92
  - `Corpus/OpenScore-LiederCorpus/Holmès,_Augusta_Mary_Anne/Les_Heures/4_L’Heure_d’Azur/` (4 files): Roman numeral analysis of Augusta Holmès, Les Heures - No.4: L'Heure d'Azur, from the OpenScore score.
93
  Format: directory of 5 files, 102,632 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5001674.mscz, score.mxl
94
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
95
- Reference result: roman_numeral_analysis_file
96
- Reference source: analysis.txt (RomanText), 59 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
97
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
98
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
99
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5001674.mscz (MuseScore source of the same score)
@@ -101,8 +81,6 @@ Units:
101
  - `Corpus/OpenScore-LiederCorpus/Jaëll,_Marie/4_Mélodies/1_À_toi/` (4 files): Roman numeral analysis of Marie Jaëll, 4 Mélodies - No.1: A toi, from the OpenScore score.
102
  Format: directory of 5 files, 67,483 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5834392.mscz, score.mxl
103
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
104
- Reference result: roman_numeral_analysis_file
105
- Reference source: analysis.txt (RomanText), 20 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students at Cornell University.'
106
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
107
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
108
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5834392.mscz (MuseScore source of the same score)
@@ -110,8 +88,6 @@ Units:
110
  - `Corpus/OpenScore-LiederCorpus/Kinkel,_Johanna/6_Lieder,_Op.10/1_Nachgefühl/` (4 files): Roman numeral analysis of Johanna Kinkel (originally published under the name J. Mathieux), 6 Lieder, Op.10 - No.1: Nachgefühl, from the OpenScore score.
111
  Format: directory of 5 files, 40,546 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6177220.mscz, score.mxl
112
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
113
- Reference result: roman_numeral_analysis_file
114
- Reference source: analysis.txt (RomanText), 8 measure lines; header line 3: 'Analyst: Anonymous contributor via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/'; Proofreader: Mark Gotham
115
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
116
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
117
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6177220.mscz (MuseScore source of the same score)
@@ -119,8 +95,6 @@ Units:
119
  - `Corpus/OpenScore-LiederCorpus/Kralik,_Mathilde/Blumenlieder/1_Maiglöckchen/` (4 files): Roman numeral analysis of Mathilde Kralik (Mathilde Aloisia Kralik von Meyrswalden), Blumenlieder - No.1: Maiglöckchen, from the OpenScore score.
120
  Format: directory of 5 files, 53,312 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6165152.mscz, score.mxl
121
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
122
- Reference result: roman_numeral_analysis_file
123
- Reference source: analysis.txt (RomanText), 22 measure lines; header line 3: 'Analyst: Lauren Irschick and Mark Gotham'
124
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
125
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
126
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6165152.mscz (MuseScore source of the same score)
@@ -128,8 +102,6 @@ Units:
128
  - `Corpus/OpenScore-LiederCorpus/Lang,_Josephine/3_Lieder,_Op.8/1_Schmetterling/` (4 files): Roman numeral analysis of Josephine Lang, 3 Lieder, Op.8 - No.1 : Schmetterling, from the OpenScore score.
129
  Format: directory of 5 files, 87,597 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6088572.mscz, score.mxl
130
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
131
- Reference result: roman_numeral_analysis_file
132
- Reference source: analysis.txt (RomanText), 54 measure lines; header line 3: 'Analyst: Anonymous contributor via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/'; Proofreader: Mark Gotham
133
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
134
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
135
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6088572.mscz (MuseScore source of the same score)
@@ -137,8 +109,6 @@ Units:
137
  - `Corpus/OpenScore-LiederCorpus/Mayer,_Emilie/3_Lieder,_Op.7/1_Du_bist_wie_eine_Blume/` (4 files): Roman numeral analysis of Emilie Mayer, 3 Lieder, Op.7 - No.1: Du bist wie eine Blume, from the OpenScore score.
138
  Format: directory of 5 files, 65,130 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5840527.mscz, score.mxl
139
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
140
- Reference result: roman_numeral_analysis_file
141
- Reference source: analysis.txt (RomanText), 32 measure lines; header line 3: 'Analyst: S. Reenan via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/'; Proofreader: Mark Gotham
142
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
143
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
144
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5840527.mscz (MuseScore source of the same score)
@@ -146,8 +116,6 @@ Units:
146
  - `Corpus/OpenScore-LiederCorpus/Paradis,_Maria_Theresia_von/12_Lieder,_1786/05_Die_Tanne_(Sieh_Doris,_wie_vom_Mond_bestrahlt_)/` (4 files): Roman numeral analysis of Maria Theresia von Paradis, 12 Lieder, 1786 (alternative title: Zwölf Lieder auf ihrer Reise in Musik gesetzt) - No.5: Die Tanne, from the OpenScore score.
147
  Format: directory of 5 files, 39,411 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5907870.mscz, score.mxl
148
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
149
- Reference result: roman_numeral_analysis_file
150
- Reference source: analysis.txt (RomanText), 17 measure lines; header line 3: 'Analyst: Anonymous contributor via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/'; Proofreader: Mark Gotham
151
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
152
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
153
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5907870.mscz (MuseScore source of the same score)
@@ -155,8 +123,6 @@ Units:
155
  - `Corpus/OpenScore-LiederCorpus/Viardot,_Pauline/6_Mélodies,_VWV_1133-11371176/4_Sérénade_à_Rosine,_VWV_1137/` (4 files): Roman numeral analysis of Pauline Viardot, 6 Mélodies, VWV 1133-1137/1176 - No.4: Sérénade à Rosine, VWV 1137, from the OpenScore score.
156
  Format: directory of 5 files, 80,142 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5989772.mscz, score.mxl
157
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
158
- Reference result: roman_numeral_analysis_file
159
- Reference source: analysis.txt (RomanText), 38 measure lines; header line 3: 'Analyst: Anonymous contributor via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/'; Proofreader: Mark Gotham
160
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
161
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
162
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5989772.mscz (MuseScore source of the same score)
@@ -164,8 +130,6 @@ Units:
164
  - `Corpus/OpenScore-LiederCorpus/White,_Maude_Valérie/3_Little_Songs/1_When_the_Swallows_Homeward_Fly/` (4 files): Roman numeral analysis of Maude Valérie White, 3 Little Songs - No.1: When the Swallows Homeward Fly, from the OpenScore score.
165
  Format: directory of 5 files, 69,796 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6202466.mscz, score.mxl
166
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
167
- Reference result: roman_numeral_analysis_file
168
- Reference source: analysis.txt (RomanText), 29 measure lines; header line 3: 'Analyst: Anonymous contributor via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/'; Proofreader: Mark Gotham
169
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
170
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
171
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6202466.mscz (MuseScore source of the same score)
@@ -173,8 +137,6 @@ Units:
173
  - `Corpus/OpenScore-LiederCorpus/Wolf,_Hugo/Eichendorff-Lieder/04_Das_Ständchen/` (5 files): Roman numeral analysis of Hugo Wolf, Eichendorff-Lieder - No.4: Das Ständchen, from the OpenScore score.
174
  Format: directory of 6 files, 614,748 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, analysis_onscore.xml, lc5024834.mscz, score.mxl
175
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
176
- Reference result: roman_numeral_analysis_file
177
- Reference source: analysis.txt (RomanText), 53 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students at Cornell University.'
178
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez and analysis_onscore.xml; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
179
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
180
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5024834.mscz (MuseScore source of the same score); analysis_onscore.xml (MusicXML with the analysis written on the score, hide with analysis.txt)
@@ -182,8 +144,6 @@ Units:
182
  - `Corpus/OpenScore-LiederCorpus/Zumsteeg,_Emilie/6_Lieder,_Op.4/1_Die_Kapelle/` (4 files): Roman numeral analysis of Emilie Zumsteeg, 6 Lieder, Op.4 - No.1: Die Kapelle, from the OpenScore score.
183
  Format: directory of 5 files, 58,398 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6159729.mscz, score.mxl
184
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
185
- Reference result: roman_numeral_analysis_file
186
- Reference source: analysis.txt (RomanText), 21 measure lines; header line 3: 'Analyst: Anna Rose Nelson via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/'; Proofreader: Mark Gotham
187
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
188
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
189
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6159729.mscz (MuseScore source of the same score)
@@ -191,8 +151,6 @@ Units:
191
  - `Corpus/OpenScore-LiederCorpus/Bonis,_Mel/_/Allons_prier!/` (4 files): Roman numeral analysis of Mélanie Bonis (1858 - 1937), Allons prier!, from the OpenScore score.
192
  Format: directory of 5 files, 69,565 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6635424.mscz, score.mxl
193
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
194
- Reference result: roman_numeral_analysis_file
195
- Reference source: analysis.txt (RomanText), 54 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
196
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
197
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
198
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6635424.mscz (MuseScore source of the same score)
@@ -200,8 +158,6 @@ Units:
200
  - `Corpus/OpenScore-LiederCorpus/Franz,_Robert/6_Gesänge,_Op.14/5_Liebesfrühling/` (4 files): Roman numeral analysis of Robert Franz, 6 Gesänge, Op. 14 - 5: Liebesfrühling, from the OpenScore score.
201
  Format: directory of 5 files, 57,454 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4976682.mscz, score.mxl
202
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
203
- Reference result: roman_numeral_analysis_file
204
- Reference source: analysis.txt (RomanText), 25 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
205
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
206
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
207
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4976682.mscz (MuseScore source of the same score)
@@ -209,8 +165,6 @@ Units:
209
  - `Corpus/OpenScore-LiederCorpus/Brahms,_Johannes/7_Lieder,_Op.48/3_Liebesklage_des_Mädchens/` (4 files): Roman numeral analysis of Johannes Brahms, 7 Lieder, Op. 48 - 3: Liebesklage des Mädchens, from the OpenScore score.
210
  Format: directory of 5 files, 76,751 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4946239.mscz, score.mxl
211
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
212
- Reference result: roman_numeral_analysis_file
213
- Reference source: analysis.txt (RomanText), 15 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
214
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
215
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
216
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4946239.mscz (MuseScore source of the same score)
@@ -218,8 +172,6 @@ Units:
218
  - `Corpus/OpenScore-LiederCorpus/Hensel,_Fanny_(Mendelssohn)/6_Lieder,_Op.1/1_Schwanenlied/` (4 files): Roman numeral analysis of Fanny (Mendelssohn) Hensel, 6 Lieder, Op.1 - No.1: Schwanenlied, from the OpenScore score.
219
  Format: directory of 5 files, 82,949 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5100543.mscz, score.mxl
220
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
221
- Reference result: roman_numeral_analysis_file
222
- Reference source: analysis.txt (RomanText), 29 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
223
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
224
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
225
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5100543.mscz (MuseScore source of the same score)
@@ -227,8 +179,6 @@ Units:
227
  - `Corpus/OpenScore-LiederCorpus/Hensel,_Fanny_(Mendelssohn)/6_Lieder,_Op.9/6_Die_Mainacht/` (4 files): Roman numeral analysis of Fanny (Mendelssohn) Hensel, 6 Lieder, Op.9 - No.6: Die Mainacht, from the OpenScore score.
228
  Format: directory of 5 files, 45,214 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4986030.mscz, score.mxl
229
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
230
- Reference result: roman_numeral_analysis_file
231
- Reference source: analysis.txt (RomanText), 16 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
232
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
233
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
234
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4986030.mscz (MuseScore source of the same score)
@@ -236,8 +186,6 @@ Units:
236
  - `Corpus/OpenScore-LiederCorpus/Hensel,_Fanny_(Mendelssohn)/5_Lieder,_Op.10/3_Abendbild/` (4 files): Roman numeral analysis of Fanny (Mendelssohn) Hensel, 5 Lieder, Op.10 - No.3: Abendbild, from the OpenScore score.
237
  Format: directory of 5 files, 65,220 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4919798.mscz, score.mxl
238
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
239
- Reference result: roman_numeral_analysis_file
240
- Reference source: analysis.txt (RomanText), 43 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
241
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
242
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
243
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4919798.mscz (MuseScore source of the same score)
@@ -245,8 +193,6 @@ Units:
245
  - `Corpus/OpenScore-LiederCorpus/Reichardt,_Louise/Zwölf_Deutsche_und_Italiänische_Romantische_Gesänge/01_Frühlingslied/` (5 files): Roman numeral analysis of Louise Reichardt, 12 Deutsche und Italiänische Romantische Gesänge - No.1: Frühlingslied, from the OpenScore score.
246
  Format: directory of 6 files, 74,908 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5067312-adjusted.mscz, lc5067312.mscz, score.mxl
247
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
248
- Reference result: roman_numeral_analysis_file
249
- Reference source: analysis.txt (RomanText), 18 measure lines; header line 3: 'Analyst: E. Marvin via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/', line 4: 'Proofreader: Mark Gotham'
250
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
251
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
252
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5067312-adjusted.mscz (MuseScore source of the same score); lc5067312.mscz (MuseScore source of the same score)
@@ -254,8 +200,6 @@ Units:
254
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Schwanengesang,_D.957/04_Ständchen/` (4 files): Roman numeral analysis of Franz Schubert, Schwanengesang - No.4: Ständchen, from the OpenScore score.
255
  Format: directory of 5 files, 63,476 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5004835.mscz, score.mxl
256
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
257
- Reference result: roman_numeral_analysis_file
258
- Reference source: analysis.txt (RomanText), 56 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
259
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
260
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
261
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5004835.mscz (MuseScore source of the same score)
@@ -263,8 +207,6 @@ Units:
263
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Schwanengesang,_D.957/08_Der_Atlas/` (4 files): Roman numeral analysis of Franz Schubert, Schwanengesang - No.8: Der Atlas, from the OpenScore score.
264
  Format: directory of 5 files, 70,493 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4985985.mscz, score.mxl
265
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
266
- Reference result: roman_numeral_analysis_file
267
- Reference source: analysis.txt (RomanText), 46 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
268
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
269
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
270
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4985985.mscz (MuseScore source of the same score)
@@ -272,8 +214,6 @@ Units:
272
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Schwanengesang,_D.957/13_Der_Doppelgänger/` (4 files): Roman numeral analysis of Franz Schubert, Schwanengesang, D.957 - 13: Der Doppelgänger, from the OpenScore score.
273
  Format: directory of 5 files, 61,631 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4985964.mscz, score.mxl
274
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
275
- Reference result: roman_numeral_analysis_file
276
- Reference source: analysis.txt (RomanText), 32 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
277
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
278
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
279
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4985964.mscz (MuseScore source of the same score)
@@ -281,8 +221,6 @@ Units:
281
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Die_schöne_Müllerin,_D.795/01_Das_Wandern/` (4 files): Roman numeral analysis of Franz Schubert, Die schöne Müllerin (D.795) - No.1: Das Wandern, from the OpenScore score.
282
  Format: directory of 5 files, 39,588 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5004731.mscz, score.mxl
283
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
284
- Reference result: roman_numeral_analysis_file
285
- Reference source: analysis.txt (RomanText), 21 measure lines; header line 3: 'Analyst: Mark Gotham'
286
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
287
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
288
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5004731.mscz (MuseScore source of the same score)
@@ -290,8 +228,6 @@ Units:
290
  - `Corpus/OpenScore-LiederCorpus/Coleridge-Taylor,_Samuel/6_Sorrow_Songs,_Op.57/2_When_I_am_dead,_my_dearest/` (4 files): Roman numeral analysis of Samuel Coleridge-Taylor, 6 Sorrow Songs, Op.57 - No.2: When I am dead, my dearest, from the OpenScore score.
291
  Format: directory of 5 files, 71,245 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6189644.mscz, score.mxl
292
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
293
- Reference result: roman_numeral_analysis_file
294
- Reference source: analysis.txt (RomanText), 36 measure lines; header line 3: 'Analyst: S. Reenan', line 4: 'Proofreader: M. Gotham'
295
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
296
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
297
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6189644.mscz (MuseScore source of the same score)
@@ -299,8 +235,6 @@ Units:
299
  - `Corpus/OpenScore-LiederCorpus/Lang,_Josephine/6_Lieder,_Op.26/1_Wiegenlied/` (4 files): Roman numeral analysis of Josephine Lang, 6 Lieder, Op.26 - No.1: Wiegenlied, from the OpenScore score.
300
  Format: directory of 5 files, 104,245 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6117412.mscz, score.mxl
301
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
302
- Reference result: roman_numeral_analysis_file
303
- Reference source: analysis.txt (RomanText), 54 measure lines; header line 3: 'Analyst: Anonymous contributor via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/', line 4: 'Proofreader: Mark Gotham'
304
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
305
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
306
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6117412.mscz (MuseScore source of the same score)
@@ -308,8 +242,6 @@ Units:
308
  - `Corpus/OpenScore-LiederCorpus/Mayer,_Emilie/3_Lieder,_Op.7/2_O_lass_mich_dein_gedenken/` (4 files): Roman numeral analysis of Emilie Mayer, 3 Lieder, Op.7 - No.2: O lass mich dein gedenken, from the OpenScore score.
309
  Format: directory of 5 files, 64,953 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5840434.mscz, score.mxl
310
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
311
- Reference result: roman_numeral_analysis_file
312
- Reference source: analysis.txt (RomanText), 28 measure lines; header line 3: 'Analyst: S. Reenan via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/', line 4: 'Proofreader: Mark Gotham'
313
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
314
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
315
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5840434.mscz (MuseScore source of the same score)
@@ -317,8 +249,6 @@ Units:
317
  - `Corpus/OpenScore-LiederCorpus/Wolf,_Hugo/Eichendorff-Lieder/19_Die_Nacht/` (4 files): Roman numeral analysis of Hugo Wolf, Eichendorff-Lieder - No.19: Die Nacht, from the OpenScore score.
318
  Format: directory of 5 files, 62,361 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5052801.mscz, score.mxl
319
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
320
- Reference result: roman_numeral_analysis_file
321
- Reference source: analysis.txt (RomanText), 17 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
322
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
323
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
324
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5052801.mscz (MuseScore source of the same score)
@@ -326,8 +256,6 @@ Units:
326
  - `Corpus/OpenScore-LiederCorpus/Jaëll,_Marie/4_Mélodies/2_Éternel_amour/` (4 files): Roman numeral analysis of Marie Jaëll, 4 Mélodies - No.2: Eternel amour, from the OpenScore score.
327
  Format: directory of 5 files, 97,640 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5837811.mscz, score.mxl
328
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
329
- Reference result: roman_numeral_analysis_file
330
- Reference source: analysis.txt (RomanText), 31 measure lines; header line 3: 'Analyst: Kate Sekula via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/', line 4: 'Proofreader: Mark Gotham'
331
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
332
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
333
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5837811.mscz (MuseScore source of the same score)
@@ -335,8 +263,6 @@ Units:
335
  - `Corpus/OpenScore-LiederCorpus/Holmès,_Augusta_Mary_Anne/Contes_divins/3_Le_chemin_du_ciel/` (4 files): Roman numeral analysis of Holmès, Augusta, Contes divins - No.3: Le Chemin du Ciel, from the OpenScore score.
336
  Format: directory of 5 files, 92,864 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5904897.mscz, score.mxl
337
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
338
- Reference result: roman_numeral_analysis_file
339
- Reference source: analysis.txt (RomanText), 33 measure lines; header line 3: 'Analyst: Anonymous contributor via the 'Working in Harmony' app: https://fourscoreandmore.org/apps/working-in-harmony/', line 4: 'Proofreader: Mark Gotham'
340
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
341
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
342
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5904897.mscz (MuseScore source of the same score)
 
8
 
9
  **Self-contained:** Each unit is self-contained. 36 of 36 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
12
 
13
  **Licence (SPDX):** CC-BY-SA-4.0 AND CC0-1.0
 
19
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/01_Gute_Nacht/score.mxl`: Roman numeral analysis of Schubert, Winterreise D.911 No.1 'Gute Nacht' from the OpenScore score.
20
  Format: directory of 5 files: score.mxl (compressed MusicXML, 29,857 bytes), analysis.txt (RomanText, 1,509 bytes), plus automatic rntxt, dez and mscz
21
  Origin: human expert analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
22
  Check: hide analysis.txt (and analysis_automatic.rntxt); the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key and Roman numeral (figure plus inversion) accuracy over the score's beats; report a key-only and a chord-only score. Expert analyses legitimately differ, so calibrate the band with a second annotator's corpus before setting a pass line
23
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
24
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez; lc5015378.mscz (MuseScore source of the same score)
25
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/05_Der_Lindenbaum/` (4 files): Roman numeral analysis of Schubert, Winterreise D.911 No.5 "Der Lindenbaum" from the OpenScore score.
26
  Format: directory of 5 files, 101,599 bytes: score.mxl (compressed MusicXML), analysis.txt (RomanText, with Form: lines), analysis_automatic.rntxt, analysis_dez_format.dez, lc5016466.mscz
27
  Origin: human expert analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
28
  Check: hide analysis.txt and analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band (for example the Winterreise Dataset annotations of the same song) is measured before any pass line is set; the Form: lines give a separate section-labelling task
29
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
30
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez; lc5016466.mscz (MuseScore source of the same score)
 
32
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/13_Die_Post/` (4 files): Roman numeral analysis of Schubert, Winterreise D.911 No.13 "Die Post" from the OpenScore score.
33
  Format: directory of 5 files, 89,732 bytes: score.mxl (compressed MusicXML), analysis.txt (RomanText, with Form: lines), analysis_automatic.rntxt, analysis_dez_format.dez, lc5007176.mscz
34
  Origin: human expert analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
35
  Check: hide analysis.txt and analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band (for example the Winterreise Dataset annotations of the same song) is measured before any pass line is set; the Form: lines give a separate section-labelling task
36
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
37
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez; lc5007176.mscz (MuseScore source of the same score)
 
39
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Winterreise,_D.911/20_Der_Wegweiser/` (4 files): Roman numeral analysis of Schubert, Winterreise D.911 No.20 "Der Wegweiser" from the OpenScore score.
40
  Format: directory of 5 files, 84,589 bytes: score.mxl (compressed MusicXML), analysis.txt (RomanText), analysis_automatic.rntxt, analysis_dez_format.dez, lc5009900.mscz
41
  Origin: human expert analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
42
  Check: hide analysis.txt and analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band (for example the Winterreise Dataset annotations of the same song) is measured before any pass line is set
43
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
44
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez; lc5009900.mscz (MuseScore source of the same score)
 
46
  - `Corpus/OpenScore-LiederCorpus/Coleridge-Taylor,_Samuel/6_Sorrow_Songs,_Op.57/1_Oh_what_comes_over_the_Sea/` (4 files): Roman numeral analysis of Samuel Coleridge-Taylor, 6 Sorrow Songs, Op.57 - No.1: Oh what comes over the Sea, from the OpenScore score.
47
  Format: directory of 5 files, 53,322 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6189652.mscz, score.mxl
48
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
49
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
50
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
51
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6189652.mscz (MuseScore source of the same score)
 
53
  - `Corpus/OpenScore-LiederCorpus/Burleigh,_Harry_Thacker/5_Songs_of_Laurence_Hope/5_Till_I_Wake/` (4 files): Roman numeral analysis of Harry Thacker Burleigh, 5 Songs of Laurence Hope - No.5: Till I Wake, from the OpenScore score.
54
  Format: directory of 5 files, 77,398 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6471199.mscz, score.mxl
55
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
56
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
57
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
58
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6471199.mscz (MuseScore source of the same score)
 
60
  - `Corpus/OpenScore-LiederCorpus/Chaminade,_Cécile/_/Berceuse/` (4 files): Roman numeral analysis of Cécile Chaminade, Berceuse, from the OpenScore score.
61
  Format: directory of 5 files, 69,112 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6264559.mscz, score.mxl
62
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
63
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
64
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
65
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6264559.mscz (MuseScore source of the same score)
 
67
  - `Corpus/OpenScore-LiederCorpus/Chausson,_Ernest/7_Mélodies,_Op.2/7_Le_Colibri/` (4 files): Roman numeral analysis of Ernest Chausson, 7 Mélodies, Op.2 - 7: Le Colibri, from the OpenScore score.
68
  Format: directory of 5 files, 67,039 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5077780.mscz, score.mxl
69
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
70
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
71
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
72
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5077780.mscz (MuseScore source of the same score)
 
74
  - `Corpus/OpenScore-LiederCorpus/Holmès,_Augusta_Mary_Anne/Les_Heures/4_L’Heure_d’Azur/` (4 files): Roman numeral analysis of Augusta Holmès, Les Heures - No.4: L'Heure d'Azur, from the OpenScore score.
75
  Format: directory of 5 files, 102,632 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5001674.mscz, score.mxl
76
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
77
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
78
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
79
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5001674.mscz (MuseScore source of the same score)
 
81
  - `Corpus/OpenScore-LiederCorpus/Jaëll,_Marie/4_Mélodies/1_À_toi/` (4 files): Roman numeral analysis of Marie Jaëll, 4 Mélodies - No.1: A toi, from the OpenScore score.
82
  Format: directory of 5 files, 67,483 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5834392.mscz, score.mxl
83
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
84
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
85
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
86
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5834392.mscz (MuseScore source of the same score)
 
88
  - `Corpus/OpenScore-LiederCorpus/Kinkel,_Johanna/6_Lieder,_Op.10/1_Nachgefühl/` (4 files): Roman numeral analysis of Johanna Kinkel (originally published under the name J. Mathieux), 6 Lieder, Op.10 - No.1: Nachgefühl, from the OpenScore score.
89
  Format: directory of 5 files, 40,546 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6177220.mscz, score.mxl
90
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
91
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
92
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
93
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6177220.mscz (MuseScore source of the same score)
 
95
  - `Corpus/OpenScore-LiederCorpus/Kralik,_Mathilde/Blumenlieder/1_Maiglöckchen/` (4 files): Roman numeral analysis of Mathilde Kralik (Mathilde Aloisia Kralik von Meyrswalden), Blumenlieder - No.1: Maiglöckchen, from the OpenScore score.
96
  Format: directory of 5 files, 53,312 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6165152.mscz, score.mxl
97
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
98
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
99
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
100
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6165152.mscz (MuseScore source of the same score)
 
102
  - `Corpus/OpenScore-LiederCorpus/Lang,_Josephine/3_Lieder,_Op.8/1_Schmetterling/` (4 files): Roman numeral analysis of Josephine Lang, 3 Lieder, Op.8 - No.1 : Schmetterling, from the OpenScore score.
103
  Format: directory of 5 files, 87,597 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6088572.mscz, score.mxl
104
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
105
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
106
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
107
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6088572.mscz (MuseScore source of the same score)
 
109
  - `Corpus/OpenScore-LiederCorpus/Mayer,_Emilie/3_Lieder,_Op.7/1_Du_bist_wie_eine_Blume/` (4 files): Roman numeral analysis of Emilie Mayer, 3 Lieder, Op.7 - No.1: Du bist wie eine Blume, from the OpenScore score.
110
  Format: directory of 5 files, 65,130 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5840527.mscz, score.mxl
111
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
112
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
113
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
114
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5840527.mscz (MuseScore source of the same score)
 
116
  - `Corpus/OpenScore-LiederCorpus/Paradis,_Maria_Theresia_von/12_Lieder,_1786/05_Die_Tanne_(Sieh_Doris,_wie_vom_Mond_bestrahlt_)/` (4 files): Roman numeral analysis of Maria Theresia von Paradis, 12 Lieder, 1786 (alternative title: Zwölf Lieder auf ihrer Reise in Musik gesetzt) - No.5: Die Tanne, from the OpenScore score.
117
  Format: directory of 5 files, 39,411 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5907870.mscz, score.mxl
118
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
119
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
120
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
121
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5907870.mscz (MuseScore source of the same score)
 
123
  - `Corpus/OpenScore-LiederCorpus/Viardot,_Pauline/6_Mélodies,_VWV_1133-11371176/4_Sérénade_à_Rosine,_VWV_1137/` (4 files): Roman numeral analysis of Pauline Viardot, 6 Mélodies, VWV 1133-1137/1176 - No.4: Sérénade à Rosine, VWV 1137, from the OpenScore score.
124
  Format: directory of 5 files, 80,142 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5989772.mscz, score.mxl
125
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
126
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
127
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
128
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5989772.mscz (MuseScore source of the same score)
 
130
  - `Corpus/OpenScore-LiederCorpus/White,_Maude_Valérie/3_Little_Songs/1_When_the_Swallows_Homeward_Fly/` (4 files): Roman numeral analysis of Maude Valérie White, 3 Little Songs - No.1: When the Swallows Homeward Fly, from the OpenScore score.
131
  Format: directory of 5 files, 69,796 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6202466.mscz, score.mxl
132
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
133
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
134
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
135
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6202466.mscz (MuseScore source of the same score)
 
137
  - `Corpus/OpenScore-LiederCorpus/Wolf,_Hugo/Eichendorff-Lieder/04_Das_Ständchen/` (5 files): Roman numeral analysis of Hugo Wolf, Eichendorff-Lieder - No.4: Das Ständchen, from the OpenScore score.
138
  Format: directory of 6 files, 614,748 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, analysis_onscore.xml, lc5024834.mscz, score.mxl
139
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
140
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez and analysis_onscore.xml; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
141
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
142
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5024834.mscz (MuseScore source of the same score); analysis_onscore.xml (MusicXML with the analysis written on the score, hide with analysis.txt)
 
144
  - `Corpus/OpenScore-LiederCorpus/Zumsteeg,_Emilie/6_Lieder,_Op.4/1_Die_Kapelle/` (4 files): Roman numeral analysis of Emilie Zumsteeg, 6 Lieder, Op.4 - No.1: Die Kapelle, from the OpenScore score.
145
  Format: directory of 5 files, 58,398 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6159729.mscz, score.mxl
146
  Origin: human analysis (contributed through the Working in Harmony app and proofread by Mark Gotham) aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
147
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
148
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
149
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6159729.mscz (MuseScore source of the same score)
 
151
  - `Corpus/OpenScore-LiederCorpus/Bonis,_Mel/_/Allons_prier!/` (4 files): Roman numeral analysis of Mélanie Bonis (1858 - 1937), Allons prier!, from the OpenScore score.
152
  Format: directory of 5 files, 69,565 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6635424.mscz, score.mxl
153
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
154
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
155
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
156
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6635424.mscz (MuseScore source of the same score)
 
158
  - `Corpus/OpenScore-LiederCorpus/Franz,_Robert/6_Gesänge,_Op.14/5_Liebesfrühling/` (4 files): Roman numeral analysis of Robert Franz, 6 Gesänge, Op. 14 - 5: Liebesfrühling, from the OpenScore score.
159
  Format: directory of 5 files, 57,454 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4976682.mscz, score.mxl
160
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
161
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
162
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
163
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4976682.mscz (MuseScore source of the same score)
 
165
  - `Corpus/OpenScore-LiederCorpus/Brahms,_Johannes/7_Lieder,_Op.48/3_Liebesklage_des_Mädchens/` (4 files): Roman numeral analysis of Johannes Brahms, 7 Lieder, Op. 48 - 3: Liebesklage des Mädchens, from the OpenScore score.
166
  Format: directory of 5 files, 76,751 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4946239.mscz, score.mxl
167
  Origin: human analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
168
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set; the Form: lines give a separate section-labelling task
169
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
170
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop from the answer side); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4946239.mscz (MuseScore source of the same score)
 
172
  - `Corpus/OpenScore-LiederCorpus/Hensel,_Fanny_(Mendelssohn)/6_Lieder,_Op.1/1_Schwanenlied/` (4 files): Roman numeral analysis of Fanny (Mendelssohn) Hensel, 6 Lieder, Op.1 - No.1: Schwanenlied, from the OpenScore score.
173
  Format: directory of 5 files, 82,949 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5100543.mscz, score.mxl
174
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
175
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
176
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
177
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5100543.mscz (MuseScore source of the same score)
 
179
  - `Corpus/OpenScore-LiederCorpus/Hensel,_Fanny_(Mendelssohn)/6_Lieder,_Op.9/6_Die_Mainacht/` (4 files): Roman numeral analysis of Fanny (Mendelssohn) Hensel, 6 Lieder, Op.9 - No.6: Die Mainacht, from the OpenScore score.
180
  Format: directory of 5 files, 45,214 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4986030.mscz, score.mxl
181
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
182
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
183
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
184
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4986030.mscz (MuseScore source of the same score)
 
186
  - `Corpus/OpenScore-LiederCorpus/Hensel,_Fanny_(Mendelssohn)/5_Lieder,_Op.10/3_Abendbild/` (4 files): Roman numeral analysis of Fanny (Mendelssohn) Hensel, 5 Lieder, Op.10 - No.3: Abendbild, from the OpenScore score.
187
  Format: directory of 5 files, 65,220 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4919798.mscz, score.mxl
188
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
189
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
190
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
191
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4919798.mscz (MuseScore source of the same score)
 
193
  - `Corpus/OpenScore-LiederCorpus/Reichardt,_Louise/Zwölf_Deutsche_und_Italiänische_Romantische_Gesänge/01_Frühlingslied/` (5 files): Roman numeral analysis of Louise Reichardt, 12 Deutsche und Italiänische Romantische Gesänge - No.1: Frühlingslied, from the OpenScore score.
194
  Format: directory of 6 files, 74,908 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5067312-adjusted.mscz, lc5067312.mscz, score.mxl
195
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
 
 
196
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
197
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
198
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5067312-adjusted.mscz (MuseScore source of the same score); lc5067312.mscz (MuseScore source of the same score)
 
200
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Schwanengesang,_D.957/04_Ständchen/` (4 files): Roman numeral analysis of Franz Schubert, Schwanengesang - No.4: Ständchen, from the OpenScore score.
201
  Format: directory of 5 files, 63,476 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5004835.mscz, score.mxl
202
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
203
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
204
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
205
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5004835.mscz (MuseScore source of the same score)
 
207
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Schwanengesang,_D.957/08_Der_Atlas/` (4 files): Roman numeral analysis of Franz Schubert, Schwanengesang - No.8: Der Atlas, from the OpenScore score.
208
  Format: directory of 5 files, 70,493 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4985985.mscz, score.mxl
209
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
210
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
211
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
212
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4985985.mscz (MuseScore source of the same score)
 
214
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Schwanengesang,_D.957/13_Der_Doppelgänger/` (4 files): Roman numeral analysis of Franz Schubert, Schwanengesang, D.957 - 13: Der Doppelgänger, from the OpenScore score.
215
  Format: directory of 5 files, 61,631 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc4985964.mscz, score.mxl
216
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
217
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
218
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
219
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc4985964.mscz (MuseScore source of the same score)
 
221
  - `Corpus/OpenScore-LiederCorpus/Schubert,_Franz/Die_schöne_Müllerin,_D.795/01_Das_Wandern/` (4 files): Roman numeral analysis of Franz Schubert, Die schöne Müllerin (D.795) - No.1: Das Wandern, from the OpenScore score.
222
  Format: directory of 5 files, 39,588 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5004731.mscz, score.mxl
223
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
224
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
225
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
226
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5004731.mscz (MuseScore source of the same score)
 
228
  - `Corpus/OpenScore-LiederCorpus/Coleridge-Taylor,_Samuel/6_Sorrow_Songs,_Op.57/2_When_I_am_dead,_my_dearest/` (4 files): Roman numeral analysis of Samuel Coleridge-Taylor, 6 Sorrow Songs, Op.57 - No.2: When I am dead, my dearest, from the OpenScore score.
229
  Format: directory of 5 files, 71,245 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6189644.mscz, score.mxl
230
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
231
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
232
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
233
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6189644.mscz (MuseScore source of the same score)
 
235
  - `Corpus/OpenScore-LiederCorpus/Lang,_Josephine/6_Lieder,_Op.26/1_Wiegenlied/` (4 files): Roman numeral analysis of Josephine Lang, 6 Lieder, Op.26 - No.1: Wiegenlied, from the OpenScore score.
236
  Format: directory of 5 files, 104,245 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc6117412.mscz, score.mxl
237
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
 
 
238
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
239
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
240
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc6117412.mscz (MuseScore source of the same score)
 
242
  - `Corpus/OpenScore-LiederCorpus/Mayer,_Emilie/3_Lieder,_Op.7/2_O_lass_mich_dein_gedenken/` (4 files): Roman numeral analysis of Emilie Mayer, 3 Lieder, Op.7 - No.2: O lass mich dein gedenken, from the OpenScore score.
243
  Format: directory of 5 files, 64,953 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5840434.mscz, score.mxl
244
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
 
 
245
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
246
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
247
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5840434.mscz (MuseScore source of the same score)
 
249
  - `Corpus/OpenScore-LiederCorpus/Wolf,_Hugo/Eichendorff-Lieder/19_Die_Nacht/` (4 files): Roman numeral analysis of Hugo Wolf, Eichendorff-Lieder - No.19: Die Nacht, from the OpenScore score.
250
  Format: directory of 5 files, 62,361 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5052801.mscz, score.mxl
251
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song
 
 
252
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
253
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
254
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5052801.mscz (MuseScore source of the same score)
 
256
  - `Corpus/OpenScore-LiederCorpus/Jaëll,_Marie/4_Mélodies/2_Éternel_amour/` (4 files): Roman numeral analysis of Marie Jaëll, 4 Mélodies - No.2: Eternel amour, from the OpenScore score.
257
  Format: directory of 5 files, 97,640 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5837811.mscz, score.mxl
258
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
 
 
259
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
260
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
261
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5837811.mscz (MuseScore source of the same score)
 
263
  - `Corpus/OpenScore-LiederCorpus/Holmès,_Augusta_Mary_Anne/Contes_divins/3_Le_chemin_du_ciel/` (4 files): Roman numeral analysis of Holmès, Augusta, Contes divins - No.3: Le Chemin du Ciel, from the OpenScore score.
264
  Format: directory of 5 files, 92,864 bytes: analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez, lc5904897.mscz, score.mxl
265
  Origin: human Roman numeral analysis aligned to a CC0 crowd-sourced engraving of a public-domain song; the analysis was entered through the 'Working in Harmony' app and proofread by Mark Gotham
 
 
266
  Check: hide analysis.txt, analysis_automatic.rntxt, analysis_dez_format.dez; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
267
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
268
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); analysis_dez_format.dez (Dezrann JSON form of the analysis, hide with analysis.txt); lc5904897.mscz (MuseScore source of the same score)
campaign_files/2026-10-md-media-and-documents/when-in-rome-variations-and-grounds/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each unit is self-contained. 2 of 2 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: roman_numeral_analysis_file. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
14
 
15
  **Licence (SPDX):** CC-BY-SA-4.0 AND CC0-1.0
@@ -21,8 +19,6 @@ Units:
21
  - `Corpus/Variations_and_Grounds/Purcell,_Henry/_/Chacony_Z730/` (3 files): Roman numeral analysis of Purcell, Chacony (Chaconne) in G Minor Z730, from the When in Rome CC0 score.
22
  Format: directory of 4 files, 71,620 bytes: analysis.txt, analysis_automatic.rntxt, score.mscz, score.mxl
23
  Origin: human Roman numeral analysis aligned to a CC0 engraving made by the analyst of a public-domain Baroque ground-bass work
24
- Reference result: roman_numeral_analysis_file
25
- Reference source: analysis.txt (RomanText), 147 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students.'
26
  Check: hide analysis.txt, analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
27
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
28
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); score.mscz (MuseScore source of the same score)
@@ -30,8 +26,6 @@ Units:
30
  - `Corpus/Variations_and_Grounds/Purcell,_Henry/_/Sonata_Z807/` (3 files): Roman numeral analysis of Purcell, Purcell Sonata in G Minor Z807, from the When in Rome CC0 score.
31
  Format: directory of 4 files, 69,439 bytes: analysis.txt, analysis_automatic.rntxt, score.mscz, score.mxl
32
  Origin: human Roman numeral analysis aligned to a CC0 engraving made by the analyst of a public-domain Baroque ground-bass work
33
- Reference result: roman_numeral_analysis_file
34
- Reference source: analysis.txt (RomanText), 222 measure lines; header line 3: 'Analyst: Mark Gotham, manually, through judicious automation, and in collaboration with colleagues and students at Cornell University'
35
  Check: hide analysis.txt, analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
36
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
37
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); score.mscz (MuseScore source of the same score)
 
8
 
9
  **Self-contained:** Each unit is self-contained. 2 of 2 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
12
 
13
  **Licence (SPDX):** CC-BY-SA-4.0 AND CC0-1.0
 
19
  - `Corpus/Variations_and_Grounds/Purcell,_Henry/_/Chacony_Z730/` (3 files): Roman numeral analysis of Purcell, Chacony (Chaconne) in G Minor Z730, from the When in Rome CC0 score.
20
  Format: directory of 4 files, 71,620 bytes: analysis.txt, analysis_automatic.rntxt, score.mscz, score.mxl
21
  Origin: human Roman numeral analysis aligned to a CC0 engraving made by the analyst of a public-domain Baroque ground-bass work
 
 
22
  Check: hide analysis.txt, analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
23
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
24
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); score.mscz (MuseScore source of the same score)
 
26
  - `Corpus/Variations_and_Grounds/Purcell,_Henry/_/Sonata_Z807/` (3 files): Roman numeral analysis of Purcell, Purcell Sonata in G Minor Z807, from the When in Rome CC0 score.
27
  Format: directory of 4 files, 69,439 bytes: analysis.txt, analysis_automatic.rntxt, score.mscz, score.mxl
28
  Origin: human Roman numeral analysis aligned to a CC0 engraving made by the analyst of a public-domain Baroque ground-bass work
 
 
29
  Check: hide analysis.txt, analysis_automatic.rntxt; the agent writes a RomanText analysis of score.mxl; parse both with music21 and score per beat: key accuracy and Roman numeral (figure plus inversion) accuracy over the score's beats, reported separately; expert analyses legitimately differ, so a second-annotator band is measured on a calibration subset before any pass line is set
30
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat; verovio renders the score headless when a picture is needed
31
  Companions (not included): analysis_automatic.rntxt (AugmentedNet machine analysis, hide with the answer); score.mscz (MuseScore source of the same score)
campaign_files/2026-10-md-media-and-documents/when-in-rome-wtc1/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: roman_numeral_analysis_file. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,8 +19,6 @@ Files:
21
  - `Corpus/Keyboard_Other/Bach,_Johann_Sebastian/The_Well-Tempered_Clavier_I/01/score.mxl`: Roman numeral analysis of J.S. Bach, Well-Tempered Clavier I, Prelude No.1 in C major.
22
  Format: directory of 3 files: score.mxl (8,831 bytes), analysis.txt (RomanText, 658 bytes), analysis_automatic.rntxt
23
  Origin: human expert analysis of a public-domain keyboard work; score file origin not stated upstream
24
- Reference result: roman_numeral_analysis_file
25
- Reference source: analysis.txt (RomanText) by Mark Gotham
26
  Check: hide analysis.txt and analysis_automatic.rntxt; the agent writes RomanText for score.mxl; music21 aligns both by measure and beat; per-beat key and Roman numeral accuracy. The C major prelude is a one-chord-per-bar texture, so the answer is close to unique; this makes it a clean low-ambiguity item
27
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
28
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop)
 
8
 
9
  **Self-contained:** The file is self-contained. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `Corpus/Keyboard_Other/Bach,_Johann_Sebastian/The_Well-Tempered_Clavier_I/01/score.mxl`: Roman numeral analysis of J.S. Bach, Well-Tempered Clavier I, Prelude No.1 in C major.
20
  Format: directory of 3 files: score.mxl (8,831 bytes), analysis.txt (RomanText, 658 bytes), analysis_automatic.rntxt
21
  Origin: human expert analysis of a public-domain keyboard work; score file origin not stated upstream
 
 
22
  Check: hide analysis.txt and analysis_automatic.rntxt; the agent writes RomanText for score.mxl; music21 aligns both by measure and beat; per-beat key and Roman numeral accuracy. The C major prelude is a one-chord-per-bar texture, so the answer is close to unique; this makes it a clean low-ambiguity item
23
  Tools: music21 (Python) parses MusicXML (.mxl) and RomanText and aligns labels by measure and beat
24
  Companions (not included): analysis_automatic.rntxt (machine analysis, drop)
campaign_files/2026-10-md-media-and-documents/will-hays-citizenship-1919/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Orthographic transcription of this recording on Wikisource (about 690 words), linked from the page's "Listen to this text" player. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent the audio file only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; the Wikisource page furniture (header, notes, licence box, footer) is removed and only the work's text is the reference; case and punctuation folded on both sides; the recording is an early acoustic phonograph disc with surface noise, so it is a hard degraded-audio item.
14
 
15
  **Licence (SPDX):** LicenseRef-PD-US-pre1931
@@ -21,7 +19,5 @@ Files:
21
  - `wikipedia/commons/5/52/Citizenship_(Will_H._Hays).ogg`: Will H. Hays, "Citizenship" (1919), phonograph recording from the Library of Congress with the Wikisource transcription of the recording.
22
  Format: Ogg Vorbis, 1,860,231 bytes, one short speech by Will H. Hays on an acoustic phonograph recording
23
  Origin: speech recorded on phonograph disc in 1918 to 1920 for the Nation's Forum series, held by the Library of Congress; text transcribed by Wikisource volunteers
24
- Reference result: orthographic transcription of this recording on Wikisource (about 690 words), linked from the page's "Listen to this text" player
25
- Reference source: wikisource/Citizenship.html, the speech body under the title and speaker line
26
  Check: give the agent the audio file only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; the Wikisource page furniture (header, notes, licence box, footer) is removed and only the work's text is the reference; case and punctuation folded on both sides; the recording is an early acoustic phonograph disc with surface noise, so it is a hard degraded-audio item
27
  Companions (not included): wikisource/Citizenship.html; File_Citizenship_(Will_H._Hays).ogg.html
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent the audio file only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; the Wikisource page furniture (header, notes, licence box, footer) is removed and only the work's text is the reference; case and punctuation folded on both sides; the recording is an early acoustic phonograph disc with surface noise, so it is a hard degraded-audio item.
12
 
13
  **Licence (SPDX):** LicenseRef-PD-US-pre1931
 
19
  - `wikipedia/commons/5/52/Citizenship_(Will_H._Hays).ogg`: Will H. Hays, "Citizenship" (1919), phonograph recording from the Library of Congress with the Wikisource transcription of the recording.
20
  Format: Ogg Vorbis, 1,860,231 bytes, one short speech by Will H. Hays on an acoustic phonograph recording
21
  Origin: speech recorded on phonograph disc in 1918 to 1920 for the Nation's Forum series, held by the Library of Congress; text transcribed by Wikisource volunteers
 
 
22
  Check: give the agent the audio file only; it writes the transcript; score word error rate (WER) against the reference words, reported as a number with no pass mark; the Wikisource page furniture (header, notes, licence box, footer) is removed and only the work's text is the reference; case and punctuation folded on both sides; the recording is an early acoustic phonograph disc with surface noise, so it is a hard degraded-audio item
23
  Companions (not included): wikisource/Citizenship.html; File_Citizenship_(Will_H._Hays).ogg.html
campaign_files/2026-10-md-media-and-documents/witiko-markdown-table-captions/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is self-contained. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Expected renderer token log for one Markdown input. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** the markdown TeX package run by a TeX engine (lualatex, or plain TeX and ConTeXt with the same setup) on the command line; the test prints the renderer token stream that the expected output records
14
 
15
  **Licence (SPDX):** LPPL-1.3c AND GPL-2.0-or-later
@@ -21,8 +19,6 @@ Files:
21
  - `tests/testfiles/unit/lunamark-markdown/table-captions.test`: Witiko/markdown unit test: pipe tables with captions converted to TeX renderer calls.
22
  Format: test file: TeX setup lines, a line "<<<", the Markdown input, a line ">>>", then the expected log of renderer tokens (BEGIN table, caption, alignment, cells)
23
  Origin: unit test written by the Witiko/markdown maintainers (lunamark-markdown group); the expected token output is specific to this package
24
- Reference result: expected renderer token log for one Markdown input
25
- Reference source: the lines after ">>>" in table-captions.test
26
  Check: split the test file at the "<<<" and ">>>" lines as read_testfile in tests/test.py does; show the agent the setup line and the Markdown input; the agent writes the renderer token log the package prints; add a final newline to the expected text when missing and pass when Python difflib.context_diff of expected and produced lines is empty
27
  Tools: the markdown TeX package run by a TeX engine (lualatex, or plain TeX and ConTeXt with the same setup) on the command line; the test prints the renderer token stream that the expected output records
28
  Companions (not included): tests/test.py; tests/README.md; LICENSE
 
8
 
9
  **Self-contained:** The file is self-contained. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** the markdown TeX package run by a TeX engine (lualatex, or plain TeX and ConTeXt with the same setup) on the command line; the test prints the renderer token stream that the expected output records
12
 
13
  **Licence (SPDX):** LPPL-1.3c AND GPL-2.0-or-later
 
19
  - `tests/testfiles/unit/lunamark-markdown/table-captions.test`: Witiko/markdown unit test: pipe tables with captions converted to TeX renderer calls.
20
  Format: test file: TeX setup lines, a line "<<<", the Markdown input, a line ">>>", then the expected log of renderer tokens (BEGIN table, caption, alignment, cells)
21
  Origin: unit test written by the Witiko/markdown maintainers (lunamark-markdown group); the expected token output is specific to this package
 
 
22
  Check: split the test file at the "<<<" and ">>>" lines as read_testfile in tests/test.py does; show the agent the setup line and the Markdown input; the agent writes the renderer token log the package prints; add a final newline to the expected text when missing and pass when Python difflib.context_diff of expected and produced lines is empty
23
  Tools: the markdown TeX package run by a TeX engine (lualatex, or plain TeX and ConTeXt with the same setup) on the command line; the test prints the renderer token stream that the expected output records
24
  Companions (not included): tests/test.py; tests/README.md; LICENSE
campaign_files/2026-10-md-media-and-documents/wspr-breqn-testfiles/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether the file needs other upstream files is not stated. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Expected normalised TeX logs (.tlg) with box dumps, one per engine. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** l3build check condition (headless LaTeX run, log normalised and diffed against the .tlg of each engine)
14
 
15
  **Licence (SPDX):** LPPL-1.3c+
@@ -21,8 +19,6 @@ Files:
21
  - `testfiles/condition.lvt`: Breqn l3build test condition: six dmath displays with \condition and \condition* variants, checked through the box log of the broken equations.
22
  Format: LaTeX test file (.lvt) plus expected logs (.tlg)
23
  Origin: regression test of the package maintainers; the .tlg files are the accepted normalised logs with \showoutput box dumps
24
- Reference result: expected normalised TeX logs (.tlg) with box dumps, one per engine
25
- Reference source: testfiles/condition.tlg (pdfTeX), condition.luatex.tlg, condition.xetex.tlg, saved by the maintainers from checked runs
26
  Check: run the .lvt with l3build on each engine and compare the normalised .log with that engine's .tlg line by line; hide the .tlg files when posing the task
27
  Tools: l3build check condition (headless LaTeX run, log normalised and diffed against the .tlg of each engine)
28
  Companions (not included): testfiles/condition.tlg; testfiles/condition.luatex.tlg; testfiles/condition.xetex.tlg; build.lua; README.md; testfiles/support/breqn-test.cls
 
8
 
9
  **Self-contained:** Whether the file needs other upstream files is not stated. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** l3build check condition (headless LaTeX run, log normalised and diffed against the .tlg of each engine)
12
 
13
  **Licence (SPDX):** LPPL-1.3c+
 
19
  - `testfiles/condition.lvt`: Breqn l3build test condition: six dmath displays with \condition and \condition* variants, checked through the box log of the broken equations.
20
  Format: LaTeX test file (.lvt) plus expected logs (.tlg)
21
  Origin: regression test of the package maintainers; the .tlg files are the accepted normalised logs with \showoutput box dumps
 
 
22
  Check: run the .lvt with l3build on each engine and compare the normalised .log with that engine's .tlg line by line; hide the .tlg files when posing the task
23
  Tools: l3build check condition (headless LaTeX run, log normalised and diffed against the .tlg of each engine)
24
  Companions (not included): testfiles/condition.tlg; testfiles/condition.luatex.tlg; testfiles/condition.xetex.tlg; build.lua; README.md; testfiles/support/breqn-test.cls
campaign_files/2026-10-md-media-and-documents/wuffs-test-data/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 5 of 5 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Reference decoding of an animation (all frames and their timings); reference decoding (same pixels as the unit, stored losslessly in another format). Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Tools are named in the check written per file in the notes.
14
 
15
  **Licence (SPDX):** MIT AND Apache-2.0
@@ -21,40 +19,30 @@ Files:
21
  - `test/data/animated-red-blue.gif`: Wuffs test data: animated-red-blue.gif with its NIA reference animation (4 frames, pixels plus cumulative durations).
22
  Format: animated GIF (64x48, 4 frames, 2,913 bytes) plus the NIA reference (Wuffs' uncompressed animation format: bn4 BGRA pixels and cumulative frame durations)
23
  Origin: original synthetic animation by Nigel Tao shipped in Wuffs' test data; test/data/README.md lines 52 to 54: "animated-red-blue.gif is an original animation by Nigel Tao ... animated-red-blue.nia ... were then generated by example/convert-to-nia"
24
- Reference result: reference decoding of an animation (all frames and their timings)
25
- Reference source: test/data/animated-red-blue.nia: every frame's pixels (BGRA, non-premultiplied, 8 bits) and the cumulative display duration of each frame in flicks; checked this round: the 4 NIA frames equal FFmpeg's bgra decoding of the GIF byte for byte
26
  Check: hide animated-red-blue.nia; the agent decodes every frame of animated-red-blue.gif (any tool) and reports the frame pixels and timings; pass when the frame count, every frame's RGBA pixels and every cumulative frame duration equal the NIA reference (doc/spec/nie-spec.md defines the layout)
27
  Companions (not included): test/data/animated-red-blue.nia; doc/spec/nie-spec.md; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
28
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
29
  - `test/data/hippopotamus.interlaced.png`: Wuffs test data: Adam7-interlaced PNG of a hippopotamus figurine with its PPM reference.
30
  Format: PNG, Adam7 interlaced (36x28 RGB) plus the reference as binary PPM (P6)
31
  Origin: crop of a museum photo shipped as Wuffs test data; test/data/README.md lines 109 to 112: "hippopotamus.* are various encodings of a cropping of a photo of \"Hippopotamus (William)\", held by the Metropolitan Museum of Art. www.metmuseum.org ... lists that image as in the public domain."
32
- Reference result: reference decoding (same pixels as the unit, stored losslessly in another format)
33
- Reference source: test/data/hippopotamus.ppm: the 36x28 RGB pixels; checked this round: FFmpeg's rgb24 decoding of the interlaced PNG equals the PPM byte for byte
34
  Check: hide hippopotamus.ppm; the agent decodes the Adam7-interlaced hippopotamus.interlaced.png to 8-bit RGB (any tool); pass when every pixel equals hippopotamus.ppm, as Wuffs' checksum list records both files with the same CRC-32 of their decoded pixels
35
  Companions (not included): test/data/hippopotamus.ppm; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
36
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
37
  - `test/data/harvesters.lossless.webp`: Wuffs test data: lossless WebP of Bruegel's The Harvesters with its PNG reference.
38
  Format: WebP lossless (VP8L) plus the reference PNG (1,000,735 pixels RGBA decoded, 4,002,940 bytes)
39
  Origin: photo of a public-domain painting shipped as Wuffs test data; test/data/README.md lines 82 to 85: "harvesters.* are various encodings of a photo of \"The Harvesters\" by Pieter Bruegel the Elder, held by the Metropolitan Museum of Art. www.metmuseum.org ... lists that image as in the public domain."
40
- Reference result: reference decoding (same pixels as the unit, stored losslessly in another format)
41
- Reference source: test/data/harvesters.png: the same pixels stored as PNG; checked this round: FFmpeg's rgba decodings of the WebP and the PNG are identical (4,002,940 bytes)
42
  Check: hide test/data/harvesters.png; the agent decodes test/data/harvesters.lossless.webp to 8-bit RGBA (any tool); pass when every pixel equals the decoding of test/data/harvesters.png, as Wuffs' checksum list records both files with the same CRC-32 of their decoded pixels
43
  Companions (not included): test/data/harvesters.png; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
44
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
45
  - `test/data/bricks-dither.gif`: Wuffs test data: dithered palette GIF of toy bricks with its PNG reference.
46
  Format: GIF (160x120, 256-colour palette, dithered) plus the reference PNG (palette)
47
  Origin: photo by a Wuffs author shipped as test data; test/data/README.md lines 60 to 61: "bricks-* are various encodings of an original photo by Nigel Tao <nigeltao@golang.org>."
48
- Reference result: reference decoding (same pixels as the unit, stored losslessly in another format)
49
- Reference source: test/data/bricks-dither.png: the same 160x120 pixels; checked this round: FFmpeg's rgba decodings of the first GIF frame and of the PNG are identical (76,800 bytes)
50
  Check: hide test/data/bricks-dither.png; the agent decodes test/data/bricks-dither.gif to 8-bit RGBA (any tool); pass when every pixel equals the decoding of test/data/bricks-dither.png, as Wuffs' checksum list records both files with the same CRC-32 of their decoded pixels
51
  Companions (not included): test/data/bricks-dither.png; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
52
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
53
  - `test/data/bricks-color.qoi`: Wuffs test data: QOI image of toy bricks with its PNG reference.
54
  Format: QOI (Quite OK Image, 160x120 RGB) plus the reference PNG
55
  Origin: photo by a Wuffs author shipped as test data; test/data/README.md lines 60 to 61: "bricks-* are various encodings of an original photo by Nigel Tao <nigeltao@golang.org>."
56
- Reference result: reference decoding (same pixels as the unit, stored losslessly in another format)
57
- Reference source: test/data/bricks-color.png: the same 160x120 RGB pixels; checked this round: FFmpeg's rgba decodings of the QOI and the PNG have the same SHA-256
58
  Check: hide test/data/bricks-color.png; the agent decodes test/data/bricks-color.qoi to 8-bit RGBA (any tool); pass when every pixel equals the decoding of test/data/bricks-color.png, as Wuffs' checksum list records both files with the same CRC-32 of their decoded pixels
59
  Companions (not included): test/data/bricks-color.png; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
60
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 5 of 5 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Tools are named in the check written per file in the notes.
12
 
13
  **Licence (SPDX):** MIT AND Apache-2.0
 
19
  - `test/data/animated-red-blue.gif`: Wuffs test data: animated-red-blue.gif with its NIA reference animation (4 frames, pixels plus cumulative durations).
20
  Format: animated GIF (64x48, 4 frames, 2,913 bytes) plus the NIA reference (Wuffs' uncompressed animation format: bn4 BGRA pixels and cumulative frame durations)
21
  Origin: original synthetic animation by Nigel Tao shipped in Wuffs' test data; test/data/README.md lines 52 to 54: "animated-red-blue.gif is an original animation by Nigel Tao ... animated-red-blue.nia ... were then generated by example/convert-to-nia"
 
 
22
  Check: hide animated-red-blue.nia; the agent decodes every frame of animated-red-blue.gif (any tool) and reports the frame pixels and timings; pass when the frame count, every frame's RGBA pixels and every cumulative frame duration equal the NIA reference (doc/spec/nie-spec.md defines the layout)
23
  Companions (not included): test/data/animated-red-blue.nia; doc/spec/nie-spec.md; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
24
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
25
  - `test/data/hippopotamus.interlaced.png`: Wuffs test data: Adam7-interlaced PNG of a hippopotamus figurine with its PPM reference.
26
  Format: PNG, Adam7 interlaced (36x28 RGB) plus the reference as binary PPM (P6)
27
  Origin: crop of a museum photo shipped as Wuffs test data; test/data/README.md lines 109 to 112: "hippopotamus.* are various encodings of a cropping of a photo of \"Hippopotamus (William)\", held by the Metropolitan Museum of Art. www.metmuseum.org ... lists that image as in the public domain."
 
 
28
  Check: hide hippopotamus.ppm; the agent decodes the Adam7-interlaced hippopotamus.interlaced.png to 8-bit RGB (any tool); pass when every pixel equals hippopotamus.ppm, as Wuffs' checksum list records both files with the same CRC-32 of their decoded pixels
29
  Companions (not included): test/data/hippopotamus.ppm; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
30
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
31
  - `test/data/harvesters.lossless.webp`: Wuffs test data: lossless WebP of Bruegel's The Harvesters with its PNG reference.
32
  Format: WebP lossless (VP8L) plus the reference PNG (1,000,735 pixels RGBA decoded, 4,002,940 bytes)
33
  Origin: photo of a public-domain painting shipped as Wuffs test data; test/data/README.md lines 82 to 85: "harvesters.* are various encodings of a photo of \"The Harvesters\" by Pieter Bruegel the Elder, held by the Metropolitan Museum of Art. www.metmuseum.org ... lists that image as in the public domain."
 
 
34
  Check: hide test/data/harvesters.png; the agent decodes test/data/harvesters.lossless.webp to 8-bit RGBA (any tool); pass when every pixel equals the decoding of test/data/harvesters.png, as Wuffs' checksum list records both files with the same CRC-32 of their decoded pixels
35
  Companions (not included): test/data/harvesters.png; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
36
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
37
  - `test/data/bricks-dither.gif`: Wuffs test data: dithered palette GIF of toy bricks with its PNG reference.
38
  Format: GIF (160x120, 256-colour palette, dithered) plus the reference PNG (palette)
39
  Origin: photo by a Wuffs author shipped as test data; test/data/README.md lines 60 to 61: "bricks-* are various encodings of an original photo by Nigel Tao <nigeltao@golang.org>."
 
 
40
  Check: hide test/data/bricks-dither.png; the agent decodes test/data/bricks-dither.gif to 8-bit RGBA (any tool); pass when every pixel equals the decoding of test/data/bricks-dither.png, as Wuffs' checksum list records both files with the same CRC-32 of their decoded pixels
41
  Companions (not included): test/data/bricks-dither.png; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
42
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
43
  - `test/data/bricks-color.qoi`: Wuffs test data: QOI image of toy bricks with its PNG reference.
44
  Format: QOI (Quite OK Image, 160x120 RGB) plus the reference PNG
45
  Origin: photo by a Wuffs author shipped as test data; test/data/README.md lines 60 to 61: "bricks-* are various encodings of an original photo by Nigel Tao <nigeltao@golang.org>."
 
 
46
  Check: hide test/data/bricks-color.png; the agent decodes test/data/bricks-color.qoi to 8-bit RGBA (any tool); pass when every pixel equals the decoding of test/data/bricks-color.png, as Wuffs' checksum list records both files with the same CRC-32 of their decoded pixels
47
  Companions (not included): test/data/bricks-color.png; test/data/README.md; test/nia-checksums-of-data.txt; LICENSE
48
  Dependencies: FFmpeg 7.1; Wuffs (C; commit 1f513d802343ec2f8a5c6e987cd071e655d36936, script/print-nia-checksums.sh) only to regenerate the checksum list.
campaign_files/2026-10-md-media-and-documents/xlsx2csv/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: Expected CSV of an .xlsx sheet with escaped control characters. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** xlsx2csv -e escape.xlsx (command line, escape mode); stdout compared with the expected CSV after removing CR
14
 
15
  **Licence (SPDX):** MIT
@@ -21,8 +19,6 @@ Files:
21
  - `test/escape.xlsx`: xlsx2csv test: workbook with newline and tab in a cell exported to CSV in escape mode.
22
  Format: .xlsx input and one expected CSV file
23
  Origin: a small workbook and its expected CSV made by the xlsx2csv maintainer; test/run is the judge
24
- Reference result: expected CSV of an .xlsx sheet with escaped control characters
25
- Reference source: test/escape.csv
26
  Check: give the agent escape.xlsx and ask for the CSV that xlsx2csv -e writes (control characters escaped as \n and \t); remove CR from both sides and require exact equality
27
  Tools: xlsx2csv -e escape.xlsx (command line, escape mode); stdout compared with the expected CSV after removing CR
28
  Companions (not included): test/escape.csv; test/run; LICENSE.txt
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 1 of 1 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** xlsx2csv -e escape.xlsx (command line, escape mode); stdout compared with the expected CSV after removing CR
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `test/escape.xlsx`: xlsx2csv test: workbook with newline and tab in a cell exported to CSV in escape mode.
20
  Format: .xlsx input and one expected CSV file
21
  Origin: a small workbook and its expected CSV made by the xlsx2csv maintainer; test/run is the judge
 
 
22
  Check: give the agent escape.xlsx and ask for the CSV that xlsx2csv -e writes (control characters escaped as \n and \t); remove CR from both sides and require exact equality
23
  Tools: xlsx2csv -e escape.xlsx (command line, escape mode); stdout compared with the expected CSV after removing CR
24
  Companions (not included): test/escape.csv; test/run; LICENSE.txt
campaign_files/2026-10-ml-serving-configs-2/wpt-webnn-conformance/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 12 of 12 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_output_tensors. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Tools are named in the check written per file in the notes.
14
 
15
  **Licence (SPDX):** BSD-3-Clause
@@ -21,96 +19,72 @@ Files:
21
  - `webnn/conformance_tests/conv2d.https.any.js`: WebNN conformance test conv2d.https.any.js: 40 conv2d cases (NCHW and NHWC layouts, padding, strides, dilations, groups, depthwise, bias, fused activation, float32 and float16) with inline inputs, filters and expected outputs.
22
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
23
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
24
- Reference result: expected_output_tensors
25
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/conv2d.https.any.js (40 test cases, float32 and float16 variants); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
26
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js lines 494 to 520 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js lines 374 to 412), which fails when the ULP distance computed by ulpDistance (utils.js from line 414) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getPrecisionTolerance (utils.js line 74), which for conv2d adds getConv2dPrecisionTolerance (utils.js line 1145).
27
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
28
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
29
  - `webnn/conformance_tests/layer_normalization.https.any.js`: WebNN conformance test layer_normalization.https.any.js: 25 layerNormalization cases (axes, scale, bias, epsilon, float32 and float16) with inline inputs and expected outputs.
30
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
31
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
32
- Reference result: expected_output_tensors
33
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/layer_normalization.https.any.js (25 test cases, float32 and float16 variants); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
34
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js lines 494 to 520 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js lines 374 to 412), which fails when the ULP distance computed by ulpDistance (utils.js from line 414) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getLayerNormPrecisionTolerance (lines 25 to 30 of the unit).
35
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
36
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
37
  - `webnn/conformance_tests/lstm.https.any.js`: WebNN conformance test lstm.https.any.js: 28 lstm cases (forward, backward and both directions, bias, recurrent bias, peephole weight, initial states, return sequence, activations, float32 and float16) with inline inputs and expected outputs.
38
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
39
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
40
- Reference result: expected_output_tensors
41
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/lstm.https.any.js (28 test cases, float32 and float16 variants); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
42
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js lines 494 to 520 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js lines 374 to 412), which fails when the ULP distance computed by ulpDistance (utils.js from line 414) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getLstmPrecisionTolerance (lines 52 to 59 of the unit).
43
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
44
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
45
  - `webnn/conformance_tests/gemm.https.any.js`: WebNN conformance test gemm.https.any.js: 51 gemm (alpha * A x B + beta * C with aTranspose and bTranspose options) cases with inline inputs and expected outputs (float32 and float16).
46
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
47
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
48
- Reference result: expected_output_tensors
49
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/gemm.https.any.js (51 test cases, float32 and float16 variants); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
50
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js from line 495 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js line 386), which fails when the ULP distance computed by ulpDistance (utils.js line 422) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getPrecisionTolerance (utils.js line 74), which for gemm adds getGemmPrecisionTolerance (utils.js line 1089).
51
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
52
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
53
  - `webnn/conformance_tests/softmax.https.any.js`: WebNN conformance test softmax.https.any.js: 9 softmax (along an axis, 2-D to 5-D inputs) cases with inline inputs and expected outputs (float32 and float16).
54
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
55
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
56
- Reference result: expected_output_tensors
57
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/softmax.https.any.js (9 test cases, float32 and float16 variants); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
58
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js from line 495 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js line 386), which fails when the ULP distance computed by ulpDistance (utils.js line 422) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getPrecisionTolerance (utils.js line 74), which for softmax adds getSoftmaxPrecisionTolerance (utils.js line 55).
59
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
60
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
61
  - `webnn/conformance_tests/gru.https.any.js`: WebNN conformance test gru.https.any.js: 24 gru (gated recurrent unit over steps, forward, backward and both directions, reset-after option, initial hidden state, bias and recurrent bias, activations) cases with inline inputs and expected outputs (float32 and float16).
62
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
63
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
64
- Reference result: expected_output_tensors
65
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/gru.https.any.js (24 test cases, float32 and float16 variants); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
66
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js from line 495 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js line 386), which fails when the ULP distance computed by ulpDistance (utils.js line 422) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getGruPrecisionTolerance, defined in the unit itself at line 52.
67
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
68
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
69
  - `webnn/conformance_tests/conv_transpose2d.https.any.js`: WebNN conformance test conv_transpose2d.https.any.js: 42 convTranspose2d cases (2D transposed convolution with padding, strides, dilations, outputPadding, outputSizes, groups, bias and filter layouts iohw, hwoi, ohwi) with inline inputs and expected outputs (float32 and float16).
70
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
71
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
72
- Reference result: expected_output_tensors
73
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/conv_transpose2d.https.any.js (42 test cases); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
74
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getPrecisionTolerance (utils.js line 74), which for convTranspose2d calls getConv2dPrecisionTolerance (utils.js line 1145); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
75
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
76
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
77
  - `webnn/conformance_tests/batch_normalization.https.any.js`: WebNN conformance test batch_normalization.https.any.js: 24 batchNormalization cases (batch normalization with axis, epsilon, scale, bias and fused activations, 1D to 5D inputs) with inline inputs and expected outputs (float32 and float16).
78
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
79
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
80
- Reference result: expected_output_tensors
81
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/batch_normalization.https.any.js (24 test cases); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
82
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getPrecisionTolerance (utils.js line 74), default branch with operatorToleranceDict (utils.js line 3); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
83
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
84
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
85
  - `webnn/conformance_tests/l2Pool2d.https.any.js`: WebNN conformance test l2Pool2d.https.any.js: 29 l2Pool2d cases (L2 pooling over 2D windows with windowDimensions, padding, strides, dilations, roundingType, outputSizes and layouts nchw and nhwc) with inline inputs and expected outputs (float32 and float16).
86
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
87
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
88
- Reference result: expected_output_tensors
89
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/l2Pool2d.https.any.js (29 test cases); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
90
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getPrecisionTolerance (utils.js line 74), which for l2Pool2d calls getPoolingOperatorsPrecisionTolerance (utils.js line 1217); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
91
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
92
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
93
  - `webnn/conformance_tests/scatterND.https.any.js`: WebNN conformance test scatterND.https.any.js: 23 scatterND cases (scatterND updates of individual elements and whole slices by N dimensional indices, including negative indices) with inline inputs and expected outputs (float32 and float16).
94
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
95
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
96
- Reference result: expected_output_tensors
97
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/scatterND.https.any.js (23 test cases); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
98
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getZeroULPTolerance (utils.js line 51); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
99
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
100
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
101
  - `webnn/conformance_tests/triangular.https.any.js`: WebNN conformance test triangular.https.any.js: 34 triangular cases (triangular (upper or lower part of the last two dimensions with a diagonal offset) on 2D to 5D tensors) with inline inputs and expected outputs (float32 and float16).
102
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
103
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
104
- Reference result: expected_output_tensors
105
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/triangular.https.any.js (34 test cases); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
106
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getTriangularPrecisionTolerance (lines 26 to 28 of the unit); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
107
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
108
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
109
  - `webnn/conformance_tests/resample2d.https.any.js`: WebNN conformance test resample2d.https.any.js: 13 resample2d cases (2D resampling (upsampling) with scales, sizes, axes and the modes nearest-neighbor and linear) with inline inputs and expected outputs (float32).
110
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
111
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
112
- Reference result: expected_output_tensors
113
- Reference source: the expectedOutputs data of each test case written inline in webnn/conformance_tests/resample2d.https.any.js (13 test cases); vectors were computed by the WebNN working group from a reference implementation of the operator as defined in the W3C WebNN specification
114
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getPrecisionTolerance (utils.js line 74), which for resample2d calls getResample2dPrecisionTolerance (utils.js line 1332); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
115
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
116
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 12 of 12 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Tools are named in the check written per file in the notes.
12
 
13
  **Licence (SPDX):** BSD-3-Clause
 
19
  - `webnn/conformance_tests/conv2d.https.any.js`: WebNN conformance test conv2d.https.any.js: 40 conv2d cases (NCHW and NHWC layouts, padding, strides, dilations, groups, depthwise, bias, fused activation, float32 and float16) with inline inputs, filters and expected outputs.
20
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
21
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
22
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js lines 494 to 520 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js lines 374 to 412), which fails when the ULP distance computed by ulpDistance (utils.js from line 414) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getPrecisionTolerance (utils.js line 74), which for conv2d adds getConv2dPrecisionTolerance (utils.js line 1145).
23
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
24
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
25
  - `webnn/conformance_tests/layer_normalization.https.any.js`: WebNN conformance test layer_normalization.https.any.js: 25 layerNormalization cases (axes, scale, bias, epsilon, float32 and float16) with inline inputs and expected outputs.
26
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
27
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
28
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js lines 494 to 520 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js lines 374 to 412), which fails when the ULP distance computed by ulpDistance (utils.js from line 414) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getLayerNormPrecisionTolerance (lines 25 to 30 of the unit).
29
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
30
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
31
  - `webnn/conformance_tests/lstm.https.any.js`: WebNN conformance test lstm.https.any.js: 28 lstm cases (forward, backward and both directions, bias, recurrent bias, peephole weight, initial states, return sequence, activations, float32 and float16) with inline inputs and expected outputs.
32
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
33
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
34
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js lines 494 to 520 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js lines 374 to 412), which fails when the ULP distance computed by ulpDistance (utils.js from line 414) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getLstmPrecisionTolerance (lines 52 to 59 of the unit).
35
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
36
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
37
  - `webnn/conformance_tests/gemm.https.any.js`: WebNN conformance test gemm.https.any.js: 51 gemm (alpha * A x B + beta * C with aTranspose and bTranspose options) cases with inline inputs and expected outputs (float32 and float16).
38
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
39
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
40
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js from line 495 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js line 386), which fails when the ULP distance computed by ulpDistance (utils.js line 422) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getPrecisionTolerance (utils.js line 74), which for gemm adds getGemmPrecisionTolerance (utils.js line 1089).
41
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
42
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
43
  - `webnn/conformance_tests/softmax.https.any.js`: WebNN conformance test softmax.https.any.js: 9 softmax (along an axis, 2-D to 5-D inputs) cases with inline inputs and expected outputs (float32 and float16).
44
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
45
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
46
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js from line 495 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js line 386), which fails when the ULP distance computed by ulpDistance (utils.js line 422) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getPrecisionTolerance (utils.js line 74), which for softmax adds getSoftmaxPrecisionTolerance (utils.js line 55).
47
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
48
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
49
  - `webnn/conformance_tests/gru.https.any.js`: WebNN conformance test gru.https.any.js: 24 gru (gated recurrent unit over steps, forward, backward and both directions, reset-after option, initial hidden state, bias and recurrent bias, activations) cases with inline inputs and expected outputs (float32 and float16).
50
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
51
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
52
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from the function named below, and doAssert in webnn/resources/utils.js from line 495 compares element by element: metricType ULP calls assert_array_approx_equals_ulp (utils.js line 386), which fails when the ULP distance computed by ulpDistance (utils.js line 422) exceeds the tolerance; the output shape and data type must also match the expectedOutputs descriptor. Tolerance function: getGruPrecisionTolerance, defined in the unit itself at line 52.
53
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
54
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit 5ae4375c.
55
  - `webnn/conformance_tests/conv_transpose2d.https.any.js`: WebNN conformance test conv_transpose2d.https.any.js: 42 convTranspose2d cases (2D transposed convolution with padding, strides, dilations, outputPadding, outputSizes, groups, bias and filter layouts iohw, hwoi, ohwi) with inline inputs and expected outputs (float32 and float16).
56
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
57
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
58
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getPrecisionTolerance (utils.js line 74), which for convTranspose2d calls getConv2dPrecisionTolerance (utils.js line 1145); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
59
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
60
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
61
  - `webnn/conformance_tests/batch_normalization.https.any.js`: WebNN conformance test batch_normalization.https.any.js: 24 batchNormalization cases (batch normalization with axis, epsilon, scale, bias and fused activations, 1D to 5D inputs) with inline inputs and expected outputs (float32 and float16).
62
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
63
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
64
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getPrecisionTolerance (utils.js line 74), default branch with operatorToleranceDict (utils.js line 3); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
65
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
66
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
67
  - `webnn/conformance_tests/l2Pool2d.https.any.js`: WebNN conformance test l2Pool2d.https.any.js: 29 l2Pool2d cases (L2 pooling over 2D windows with windowDimensions, padding, strides, dilations, roundingType, outputSizes and layouts nchw and nhwc) with inline inputs and expected outputs (float32 and float16).
68
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
69
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
70
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getPrecisionTolerance (utils.js line 74), which for l2Pool2d calls getPoolingOperatorsPrecisionTolerance (utils.js line 1217); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
71
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
72
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
73
  - `webnn/conformance_tests/scatterND.https.any.js`: WebNN conformance test scatterND.https.any.js: 23 scatterND cases (scatterND updates of individual elements and whole slices by N dimensional indices, including negative indices) with inline inputs and expected outputs (float32 and float16).
74
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
75
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
76
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getZeroULPTolerance (utils.js line 51); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
77
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
78
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
79
  - `webnn/conformance_tests/triangular.https.any.js`: WebNN conformance test triangular.https.any.js: 34 triangular cases (triangular (upper or lower part of the last two dimensions with a diagonal offset) on 2D to 5D tensors) with inline inputs and expected outputs (float32 and float16).
80
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
81
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
82
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getTriangularPrecisionTolerance (lines 26 to 28 of the unit); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
83
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
84
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
85
  - `webnn/conformance_tests/resample2d.https.any.js`: WebNN conformance test resample2d.https.any.js: 13 resample2d cases (2D resampling (upsampling) with scales, sizes, axes and the modes nearest-neighbor and linear) with inline inputs and expected outputs (float32).
86
  Format: JavaScript test file (testharness.js, .https.any.js) with inline input and expected output arrays
87
  Origin: conformance fixture of the W3C WebNN API written by the WebNN working group
 
 
88
  Check: Build and run each graph of the file on the WebNN backend under test (buildAndExecuteGraph in utils.js) and compare each output with expectedOutputs. The harness takes the tolerance from getPrecisionTolerance (utils.js line 74), which for resample2d calls getResample2dPrecisionTolerance (utils.js line 1332); doAssert in webnn/resources/utils.js from line 493 compares element by element in ULP (assert_array_approx_equals_ulp, utils.js line 386, with ulpDistance at line 422); the output shape and data type must also match the expectedOutputs descriptor.
89
  Companions (not included): webnn/resources/utils.js; resources/testharness.js; resources/testharnessreport.js; LICENSE.md
90
  Dependencies: a browser that implements the WebNN API (navigator.ml), for example Chromium with WebNN enabled. the web-platform-tests runner (wpt run) from the same commit da1355b3.
campaign_files/2026-10-ml-serving-configs-2/xgrammar-structured-generation-test-expectations/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 9 of 9 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_test_outputs. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Tools are named in the check written per file in the notes.
14
 
15
  **Licence (SPDX):** Apache-2.0
@@ -21,72 +19,54 @@ Files:
21
  - `tests/python/test_json_schema_converter.py`: XGrammar JSON Schema to EBNF: 106 tests give a JSON schema or pydantic model, the exact EBNF grammar the converter must emit, and JSON instances it must accept or reject.
22
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
23
  Origin: expectations written by the XGrammar maintainers into the test sources
24
- Reference result: expected_test_outputs
25
- Reference source: the expected EBNF strings and the accepted and rejected instance lists inside each test, written by the maintainers
26
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_json_schema_converter.py out with the companions under tests/, and run pytest tests/python/test_json_schema_converter.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected EBNF literals and the accept or reject flags; a solver supplies the grammar or the verdicts and the hidden literals are compared. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
27
  Companions (not included): tests/conftest.py; LICENSE
28
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
29
  - `tests/python/test_grammar_matcher_ebnf.py`: XGrammar EBNF grammar matcher: 34 tests check which strings a grammar matcher accepts, the token bitmasks it fills and the jump-forward strings it emits.
30
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
31
  Origin: expectations written by the XGrammar maintainers into the test sources
32
- Reference result: expected_test_outputs
33
- Reference source: the expected accept or reject verdicts, jump-forward strings and rollback states inside each test
34
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_grammar_matcher_ebnf.py out with the companions under tests/, and run pytest tests/python/test_grammar_matcher_ebnf.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected verdicts and jump-forward strings. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance. The three hf_token_required tests (lines 370, 642, 684) load gated Llama tokenizers and are out.
35
  Companions (not included): tests/conftest.py; tests/python/tokenizer_utils.py; LICENSE
36
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
37
  - `tests/python/test_regex_converter.py`: XGrammar regex to EBNF: 29 tests give a regular expression, the exact EBNF it must convert to, and strings it must accept or reject.
38
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
39
  Origin: expectations written by the XGrammar maintainers into the test sources
40
- Reference result: expected_test_outputs
41
- Reference source: the expected EBNF strings and the accept and reject instance lists (for example regex_expected_grammar_instance at line 58)
42
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_regex_converter.py out with the companions under tests/, and run pytest tests/python/test_regex_converter.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected EBNF literals and verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance. The hf_token_required test at line 466 uses gated tokenizers and is out.
43
  Companions (not included): tests/conftest.py; tests/python/tokenizer_utils.py; LICENSE
44
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
45
  - `tests/python/test_grammar_parser.py`: XGrammar EBNF parser: 49 tests give an input grammar and the exact normalized grammar text the parser must print after its rewrite passes, or the error it must raise.
46
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
47
  Origin: expectations written by the XGrammar maintainers into the test sources
48
- Reference result: expected_test_outputs
49
- Reference source: the expected normalized grammar strings and error messages inside each test
50
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_grammar_parser.py out with the companions under tests/, and run pytest tests/python/test_grammar_parser.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected printed grammars and error messages. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
51
  Companions (not included): tests/conftest.py; LICENSE
52
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
53
  - `tests/python/test_lark.py`: XGrammar Lark front end: 158 tests give a Lark grammar, the exact EBNF it must translate to, and strings it must accept or reject.
54
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
55
  Origin: expectations written by the XGrammar maintainers into the test sources
56
- Reference result: expected_test_outputs
57
- Reference source: the expected EBNF strings and the accept and reject lists inside each test
58
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_lark.py out with the companions under tests/, and run pytest tests/python/test_lark.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected EBNF literals and verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
59
  Companions (not included): tests/conftest.py; LICENSE
60
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
61
  - `tests/python/test_structural_tag_converter.py`: XGrammar structural tags: 108 tests give a structural-tag spec (tags, sequences, JSON-schema bodies), the exact grammar it must compile to, and generations it must accept or reject.
62
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
63
  Origin: expectations written by the XGrammar maintainers into the test sources
64
- Reference result: expected_test_outputs
65
- Reference source: the expected grammar strings and accept or reject lists inside each test
66
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_structural_tag_converter.py out with the companions under tests/, and run pytest tests/python/test_structural_tag_converter.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected grammar literals and verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance. The optional Profiler loads a gated tokenizer only when a Hugging Face token is set and only times the run; it is not part of any answer.
67
  Companions (not included): tests/conftest.py; tests/python/tokenizer_utils.py; LICENSE
68
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
69
  - `tests/python/test_function_calling_converter.py`: XGrammar function-calling formats: 90 tests give tool definitions in a model's tool-call style, the exact grammar it must compile to, and tool-call outputs it must accept or reject.
70
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
71
  Origin: expectations written by the XGrammar maintainers into the test sources
72
- Reference result: expected_test_outputs
73
- Reference source: the expected grammar strings and accept or reject lists inside each test
74
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_function_calling_converter.py out with the companions under tests/, and run pytest tests/python/test_function_calling_converter.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected grammar literals and verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
75
  Companions (not included): tests/conftest.py; LICENSE
76
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
77
  - `tests/python/test_builtin_structural_tag.py`: XGrammar built-in tool-call formats: 73 tests give OpenAI-style tool lists for named model families (Llama, Qwen, DeepSeek, Kimi, MiniMax, GLM, Cohere and the Harmony format) and the tool-call texts the built-in structural tag must accept or reject.
78
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
79
  Origin: expectations written by the XGrammar maintainers into the test sources
80
- Reference result: expected_test_outputs
81
- Reference source: the expected accept or reject verdicts and the expected structural-tag grammars inside each test
82
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_builtin_structural_tag.py out with the companions under tests/, and run pytest tests/python/test_builtin_structural_tag.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected verdicts and grammar literals. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance. The optional Profiler loads a gated tokenizer only when a Hugging Face token is set and only times the run; it is not part of any answer.
83
  Companions (not included): tests/conftest.py; tests/python/tokenizer_utils.py; LICENSE
84
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
85
  - `tests/python/test_json_schema_excludes.py`: XGrammar JSON schema with excluded keys and values: 31 tests give a schema with exclusions and the instances the compiled grammar must accept or reject.
86
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
87
  Origin: expectations written by the XGrammar maintainers into the test sources
88
- Reference result: expected_test_outputs
89
- Reference source: the expected accept or reject verdicts inside each test
90
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_json_schema_excludes.py out with the companions under tests/, and run pytest tests/python/test_json_schema_excludes.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
91
  Companions (not included): tests/conftest.py; LICENSE
92
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
 
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 9 of 9 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Tools are named in the check written per file in the notes.
12
 
13
  **Licence (SPDX):** Apache-2.0
 
19
  - `tests/python/test_json_schema_converter.py`: XGrammar JSON Schema to EBNF: 106 tests give a JSON schema or pydantic model, the exact EBNF grammar the converter must emit, and JSON instances it must accept or reject.
20
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
21
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
22
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_json_schema_converter.py out with the companions under tests/, and run pytest tests/python/test_json_schema_converter.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected EBNF literals and the accept or reject flags; a solver supplies the grammar or the verdicts and the hidden literals are compared. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
23
  Companions (not included): tests/conftest.py; LICENSE
24
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
25
  - `tests/python/test_grammar_matcher_ebnf.py`: XGrammar EBNF grammar matcher: 34 tests check which strings a grammar matcher accepts, the token bitmasks it fills and the jump-forward strings it emits.
26
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
27
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
28
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_grammar_matcher_ebnf.py out with the companions under tests/, and run pytest tests/python/test_grammar_matcher_ebnf.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected verdicts and jump-forward strings. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance. The three hf_token_required tests (lines 370, 642, 684) load gated Llama tokenizers and are out.
29
  Companions (not included): tests/conftest.py; tests/python/tokenizer_utils.py; LICENSE
30
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
31
  - `tests/python/test_regex_converter.py`: XGrammar regex to EBNF: 29 tests give a regular expression, the exact EBNF it must convert to, and strings it must accept or reject.
32
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
33
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
34
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_regex_converter.py out with the companions under tests/, and run pytest tests/python/test_regex_converter.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected EBNF literals and verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance. The hf_token_required test at line 466 uses gated tokenizers and is out.
35
  Companions (not included): tests/conftest.py; tests/python/tokenizer_utils.py; LICENSE
36
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
37
  - `tests/python/test_grammar_parser.py`: XGrammar EBNF parser: 49 tests give an input grammar and the exact normalized grammar text the parser must print after its rewrite passes, or the error it must raise.
38
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
39
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
40
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_grammar_parser.py out with the companions under tests/, and run pytest tests/python/test_grammar_parser.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected printed grammars and error messages. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
41
  Companions (not included): tests/conftest.py; LICENSE
42
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
43
  - `tests/python/test_lark.py`: XGrammar Lark front end: 158 tests give a Lark grammar, the exact EBNF it must translate to, and strings it must accept or reject.
44
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
45
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
46
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_lark.py out with the companions under tests/, and run pytest tests/python/test_lark.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected EBNF literals and verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
47
  Companions (not included): tests/conftest.py; LICENSE
48
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
49
  - `tests/python/test_structural_tag_converter.py`: XGrammar structural tags: 108 tests give a structural-tag spec (tags, sequences, JSON-schema bodies), the exact grammar it must compile to, and generations it must accept or reject.
50
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
51
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
52
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_structural_tag_converter.py out with the companions under tests/, and run pytest tests/python/test_structural_tag_converter.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected grammar literals and verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance. The optional Profiler loads a gated tokenizer only when a Hugging Face token is set and only times the run; it is not part of any answer.
53
  Companions (not included): tests/conftest.py; tests/python/tokenizer_utils.py; LICENSE
54
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
55
  - `tests/python/test_function_calling_converter.py`: XGrammar function-calling formats: 90 tests give tool definitions in a model's tool-call style, the exact grammar it must compile to, and tool-call outputs it must accept or reject.
56
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
57
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
58
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_function_calling_converter.py out with the companions under tests/, and run pytest tests/python/test_function_calling_converter.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected grammar literals and verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
59
  Companions (not included): tests/conftest.py; LICENSE
60
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
61
  - `tests/python/test_builtin_structural_tag.py`: XGrammar built-in tool-call formats: 73 tests give OpenAI-style tool lists for named model families (Llama, Qwen, DeepSeek, Kimi, MiniMax, GLM, Cohere and the Harmony format) and the tool-call texts the built-in structural tag must accept or reject.
62
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
63
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
64
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_builtin_structural_tag.py out with the companions under tests/, and run pytest tests/python/test_builtin_structural_tag.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected verdicts and grammar literals. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance. The optional Profiler loads a gated tokenizer only when a Hugging Face token is set and only times the run; it is not part of any answer.
65
  Companions (not included): tests/conftest.py; tests/python/tokenizer_utils.py; LICENSE
66
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
67
  - `tests/python/test_json_schema_excludes.py`: XGrammar JSON schema with excluded keys and values: 31 tests give a schema with exclusions and the instances the compiled grammar must accept or reject.
68
  Format: pytest Python module holding grammars, JSON schemas or tool specs with the expected grammar text and the strings that must be accepted or rejected, inline
69
  Origin: expectations written by the XGrammar maintainers into the test sources
 
 
70
  Check: Install xgrammar 0.2.8 with the test extra on CPU, lay tests/python/test_json_schema_excludes.py out with the companions under tests/, and run pytest tests/python/test_json_schema_excludes.py -m "not hf_token_required" (the upstream CI selection, pyproject.toml line 168). Hide the expected verdicts. Every comparison is an exact string or boolean equality (assert ==, assert is True/False); there is no numeric tolerance.
71
  Companions (not included): tests/conftest.py; LICENSE
72
  Dependencies: xgrammar==0.2.8 (the pinned release; tests/python/test_*.py import it as xgr and xgrammar.testing); pytest, pydantic, torch>=1.10.0, transformers>=4.38.0, numpy (pyproject.toml lines 17 to 31 and the test extra); not bundled
campaign_files/2026-10-ml-serving-configs-2/xilinx-brevitas/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 10 of 10 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: aggregate_metric_reference. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Tools are named in the check written per file in the notes.
14
 
15
  **Licence (SPDX):** BSD-3-Clause
@@ -21,80 +19,60 @@ Files:
21
  - `src/brevitas_examples/imagenet_classification/ptq/RESULTS_TORCHVISION_BEST_CONFIGS.csv`: Brevitas post-training quantization of torchvision ImageNet models: best configuration per hardware feature set, with float and quantized top-1 (RESULTS_TORCHVISION_BEST_CONFIGS.csv).
22
  Format: CSV result file plus Python evaluation scripts and YAML config template
23
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
24
- Reference result: aggregate_metric_reference
25
- Reference source: RESULTS_TORCHVISION_BEST_CONFIGS.csv lines 2-181, columns `Top 1% floating point accuracy` and `Top 1% quant accuracy` (header line 1), 3 decimals; README.md lines 53-54 say the file lists, for each combination of hardware-related features, the best configuration from the full sweep RESULTS_TORCHVISION.csv
26
  Check: Pick one CSV row, run brevitas_ptq_imagenet_val (ptq_evaluate.py) at the pinned commit with that row's settings (model, target backend, scale factor type, weight, activation and bias bit widths, per-channel scaling, activation quant type, bias correction, graph equalization iterations and merge bias, GPTQ and act order, activation percentile, calibration size 1000 and batch size 64) on the ImageNet validation set, using torch 1.13.1 and brevitas at the version of the row; pass when the recomputed float and quantized top-1 match the printed values within the tolerance below. Rows made with other library versions are not comparable; the README warns results may vary across versions.
27
  Companions (not included): src/brevitas_examples/imagenet_classification/ptq/README.md; src/brevitas_examples/imagenet_classification/ptq/ptq_evaluate.py; src/brevitas_examples/imagenet_classification/ptq/ptq_common.py; src/brevitas_examples/imagenet_classification/ptq/ptq_imagenet_args.py; src/brevitas_examples/imagenet_classification/ptq/utils.py; src/brevitas_examples/imagenet_classification/ptq/config/default_template.yaml; src/brevitas_examples/imagenet_classification/ptq/benchmark/ptq_benchmark.py; src/brevitas_examples/imagenet_classification/ptq/benchmark/benchmark_template.yaml; src/brevitas_examples/imagenet_classification/utils.py; LICENSE
28
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (setup.py, requirements/requirements.txt) and torchvision pretrained weights for the 3 models in the CSV (mobilenet_v2, resnet18, vit_b_32 as named in column 1; BSD-3-Clause torchvision weights); the CSV was produced with torch 1.13.1 and brevitas 0.9.2.dev13+g96ddb35e (last two columns), so a recompute pins those versions. ImageNet-1k (ILSVRC2012) validation set, 50,000 images, gated, plus a calibration folder of 1,000 training images (ptq_imagenet_args.py line 48, --calibration-samples default 1000).
29
  - `src/brevitas_examples/imagenet_classification/ptq/RESULTS_IMGCLSMOB.csv`: Brevitas post-training quantization of the imgclsmob quant_mobilenet_v1 on ImageNet: float vs quantized top-1 (RESULTS_IMGCLSMOB.csv).
30
  Format: CSV result file plus Python evaluation scripts and an INI model config
31
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
32
- Reference result: aggregate_metric_reference
33
- Reference source: RESULTS_IMGCLSMOB.csv line 2, columns `Top 1% floating point accuracy` 73.39 and `Top 1% quant accuracy` 72.704 (header line 1)
34
  Check: Run brevitas_ptq_imagenet_val at the pinned commit for the imgclsmob quant_mobilenet_v1 with bias correction, calibration size 1000 and batch size 64, under torch 1.13.1 and the brevitas version of the row; pass when float and quantized top-1 match the printed values within the tolerances below.
35
  Companions (not included): src/brevitas_examples/imagenet_classification/ptq/README.md; src/brevitas_examples/imagenet_classification/ptq/ptq_evaluate.py; src/brevitas_examples/imagenet_classification/ptq/ptq_common.py; src/brevitas_examples/imagenet_classification/ptq/ptq_imagenet_args.py; src/brevitas_examples/imagenet_classification/ptq/utils.py; src/brevitas_examples/imagenet_classification/models/cfg/quant_mobilenet_v1.ini; src/brevitas_examples/imagenet_classification/models/README.md; src/brevitas_examples/imagenet_classification/utils.py; LICENSE
36
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d and the float MobileNet V1 weights the model config names (imgclsmob / pytorchcv, MIT per the source headers of models/mobilenetv1.py); the CSV was produced with torch 1.13.1 and brevitas 0.8.1.dev98, so a recompute pins those versions. ImageNet-1k (ILSVRC2012) validation set, 50,000 images, gated, plus a calibration folder of 1,000 training images (ptq_imagenet_args.py line 48, --calibration-samples default 1000).
37
  - `src/brevitas_examples/imagenet_classification/qat/README.md`: Brevitas quantization-aware trained ImageNet models (4-bit MobileNet V1 and ProxylessNAS Mobile14): published top-1 with released checkpoints.
38
  Format: Markdown results table plus Python evaluation script and INI model configs
39
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
40
- Reference result: aggregate_metric_reference
41
- Reference source: qat/README.md lines 10-13, column Top1: quant_mobilenet_v1_4b 70.86, quant_proxylessnas_mobile14_hadamard_4b 72.87, quant_proxylessnas_mobile14_4b 74.39, quant_proxylessnas_mobile14_4b5b 74.94
42
  Check: For one model, run `brevitas_imagenet_val --imagenet-dir <imagenet> --model <cfg name> --pretrained` (qat/README.md line 24; the checkpoint is downloaded from the release) at the pinned commit and compare top-1 with the README value within the tolerance below.
43
  Companions (not included): src/brevitas_examples/imagenet_classification/qat/imagenet_val.py; src/brevitas_examples/imagenet_classification/models/cfg/quant_mobilenet_v1_4b.ini; src/brevitas_examples/imagenet_classification/models/cfg/quant_proxylessnas_mobile14_hadamard_4b.ini; src/brevitas_examples/imagenet_classification/models/cfg/quant_proxylessnas_mobile14_4b.ini; src/brevitas_examples/imagenet_classification/models/cfg/quant_proxylessnas_mobile14_4b5b.ini; src/brevitas_examples/imagenet_classification/models/README.md; src/brevitas_examples/imagenet_classification/utils.py; LICENSE
44
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (package brevitas_examples provides the model definitions models/mobilenetv1.py and models/proxylessnas.py, MIT, adapted from imgclsmob). the 4 quantized checkpoints released on GitHub (github.com/Xilinx/brevitas/releases, tags quant_mobilenet_v1_4b-r1 and quant_proxylessnas_mobile14_*-r0, files named in README lines 10-13 with hash suffixes); release assets under the repository licence. ImageNet-1k (ILSVRC2012) validation set, 50,000 images, gated.
45
  - `src/brevitas_examples/imagenet_classification/a2q/README.md`: Brevitas accumulator-aware quantization (A2Q, A2Q+) of ResNet18 on CIFAR10: published top-1 per target accumulator width with released checkpoints.
46
  Format: Markdown results table plus Python evaluation script
47
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
48
- Reference result: aggregate_metric_reference
49
- Reference source: a2q/README.md lines 16-28, column `Top-1 Accuracy (%)`, one decimal (e.g. float_resnet18 95.0, quant_resnet18_w4a4_a2q_12b 83.9, quant_resnet18_w4a4_a2q_plus_12b 90.6)
50
  Check: Run a2q_evaluate_models.py at the pinned commit for one model name of the table with the released checkpoint on the CIFAR-10 test set, and compare top-1 with the README value within the tolerance below.
51
  Companions (not included): src/brevitas_examples/imagenet_classification/a2q/a2q_evaluate_models.py; src/brevitas_examples/imagenet_classification/a2q/utils.py; LICENSE
52
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (package brevitas_examples provides a2q/resnet.py, MIT, adapted from kuangliu pytorch-cifar, and quant.py, ep_init.py). the 11 checkpoints released on GitHub (github.com/Xilinx/brevitas/releases tag a2q_cifar10_r1, file names with hash suffixes in README lines 16-28). CIFAR-10 test set (10,000 images, University of Toronto, no licence stated on the dataset page); downloaded by torchvision at run time.
53
  - `src/brevitas_examples/bnn_pynq/README.md`: Brevitas binarized and low-bit networks for FINN/PYNQ (TFC, SFC, LFC on MNIST; CNV and ResNet18 on CIFAR10): published top-1 with pretrained checkpoints.
54
  Format: Markdown results table plus Python evaluation scripts and INI model configs
55
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
56
- Reference result: aggregate_metric_reference
57
- Reference source: bnn_pynq/README.md lines 12-23, column `Top1 accuracy`, 2 decimals with a percent sign (e.g. TFC_1W1A 93.17%, CNV_2W2A 89.03%, RESNET18_4W4A 92.61%); each cfg/*.ini also names the maintainers' evaluation log (EVAL_LOG, e.g. cfg/tfc_1w1a.ini line 4, a release asset not fetched this round)
58
  Check: For one model, run `BREVITAS_JIT=1 python bnn_pynq_train.py --evaluate --network <NAME> --pretrained` (README line 41) at the pinned commit on the test set named in the table, and compare top-1 with the README value within the tolerance below.
59
  Companions (not included): src/brevitas_examples/bnn_pynq/bnn_pynq_train.py; src/brevitas_examples/bnn_pynq/trainer.py; src/brevitas_examples/bnn_pynq/logger.py; src/brevitas_examples/bnn_pynq/cfg/tfc_1w1a.ini; src/brevitas_examples/bnn_pynq/cfg/tfc_1w2a.ini; src/brevitas_examples/bnn_pynq/cfg/tfc_2w2a.ini; src/brevitas_examples/bnn_pynq/cfg/sfc_1w1a.ini; src/brevitas_examples/bnn_pynq/cfg/sfc_1w2a.ini; src/brevitas_examples/bnn_pynq/cfg/sfc_2w2a.ini; src/brevitas_examples/bnn_pynq/cfg/lfc_1w1a.ini; src/brevitas_examples/bnn_pynq/cfg/lfc_1w2a.ini; src/brevitas_examples/bnn_pynq/cfg/cnv_1w1a.ini; src/brevitas_examples/bnn_pynq/cfg/cnv_1w2a.ini; src/brevitas_examples/bnn_pynq/cfg/cnv_2w2a.ini; src/brevitas_examples/bnn_pynq/cfg/resnet18_4w4a.ini; LICENSE
60
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (package brevitas_examples provides models/*.py). pretrained checkpoints whose URLs are in the cfg/*.ini files (GitHub releases of Xilinx/brevitas). MNIST test set (10,000 images; Yann LeCun and Corinna Cortes; licence statements vary by mirror) for the 8 FC models and CIFAR-10 test set (10,000 images; no licence stated) for the 4 CIFAR models; downloaded by torchvision at run time.
61
  - `src/brevitas_examples/speech_to_text/README.md`: Brevitas quantized QuartzNet 15x5 speech recognition (8-bit and 4-bit): published WER on LibriSpeech dev-other with released checkpoints.
62
  Format: Markdown results table plus Python evaluation script, INI and YAML configs
63
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
64
- Reference result: aggregate_metric_reference
65
- Reference source: speech_to_text/README.md lines 10-12, column `WER (Word Error Rate) on dev-other`: 11.03%, 10.98% and 12.00%
66
  Check: Prepare LibriSpeech dev-other with brevitas_quartznet_preprocess, run `brevitas_quartznet_val --data-json <json> --model <cfg name> --pretrained` (README line 33) at the pinned commit, and compare the WER with the README value within the tolerance below.
67
  Companions (not included): src/brevitas_examples/speech_to_text/quartznet_val.py; src/brevitas_examples/speech_to_text/cfg/quant_quartznet_pertensorscaling_8b.ini; src/brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_8b.ini; src/brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_4b.ini; src/brevitas_examples/speech_to_text/cfg/topology/quartznet15x5.yaml; LICENSE
68
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (package brevitas_examples provides speech_to_text/quartznet/, adapted from NVIDIA NeMo r0.9, Apache-2.0 headers). encoder and decoder checkpoints released on GitHub (tags quant_quartznet_8b-r0 and quant_quartznet_4b-r0, README lines 10-12). LibriSpeech dev-other (openslr.org SLR12, CC BY 4.0), prepared by brevitas_quartznet_preprocess (get_librispeech_data.py).
69
  - `src/brevitas_examples/papers/gguf/README.md`: Brevitas GGUF export validation: in-process perplexity vs llama.cpp llama-perplexity for 13 GGUF recipes on three Llama 3 models, with the Qronos Q2_K configs.
70
  Format: Markdown results tables plus YAML configs and Python recipe plugins
71
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
72
- Reference result: aggregate_metric_reference
73
- Reference source: papers/gguf/README.md lines 174-186, 192-204 and 210-222: Brevitas PPL and llama.cpp PPL per recipe, 2 decimals (e.g. Llama-3.2-1B Q8_0 11.81 and 11.81, Q2_K 31.58 and 30.60); lines 19-20 the Q2_K RTN and Qronos llama.cpp PPL (30.60, 16.97; 15.04, 13.16; 9.59, 7.90)
74
  Check: Export one recipe of one model to GGUF with brevitas at the pinned commit (fields custom_quantizer and export_target as in the README config section, lines 90-130, field table lines 96-98; the Qronos rows use the provided llama3-*-q2_k.yml), build wiki.test.raw as README lines 140-151 do, run `llama-perplexity -m <model.gguf> -f wiki.test.raw -c 2048 -ngl 0` (line 157, 141 chunks), and compare the llama.cpp PPL with the table value within the half-unit tolerance; the Brevitas in-process PPL is compared to the llama.cpp PPL of the same row within the published harness gap.
75
  Companions (not included): src/brevitas_examples/papers/gguf/llama3-1b-q2_k.yml; src/brevitas_examples/papers/gguf/llama3-3b-q2_k.yml; src/brevitas_examples/papers/gguf/llama3-8b-q2_k.yml; src/brevitas_examples/papers/gguf/recipes/Llama-3.2-1B.py; src/brevitas_examples/papers/gguf/recipes/Llama-3.2-3B.py; src/brevitas_examples/papers/gguf/recipes/Llama-3.1-8B.py; src/brevitas_examples/papers/gguf/recipes/common.py; LICENSE
76
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d with its llm extras (entry point brevitas_ptq_llm). llama.cpp llama-perplexity (MIT) at a pinned release. Llama 3.2 1B Instruct, Llama 3.2 3B Instruct and Llama 3.1 8B Instruct weights from Hugging Face (meta-llama, gated, Llama 3.2 and Llama 3.1 Community Licence; allowed with the gate noted). WikiText-2 raw test split (Salesforce wikitext on Hugging Face, CC BY-SA 3.0 / GFDL).
77
  - `src/brevitas_examples/papers/axe/README.md`: Brevitas accumulator-aware post-training quantization of LLMs (AXE, TMLR 2025) on Llama 3.2 1B Instruct: WikiText2 perplexity and five zero-shot accuracies per configuration.
78
  Format: Markdown results tables plus YAML sweep configs
79
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
80
- Reference result: aggregate_metric_reference
81
- Reference source: papers/axe/README.md lines 61, 78-96 and 107-117: Wiki2 PPL (1 decimal in the baseline and channelwise tables, 2 decimals in the groupwise tables) and zero-shot accuracies with 1 decimal
82
  Check: Run one configuration of the sweep (the matching entry of benchmark-llama3-channelwise.yml or benchmark-llama3-groupwise.yml, via brevitas_ptq_llm) at the pinned commit with the pinned torch, transformers and lighteval, and compare Wiki2 PPL and each zero-shot accuracy with the table row within the tolerances below.
83
  Companions (not included): src/brevitas_examples/papers/axe/benchmark-llama3-channelwise.yml; src/brevitas_examples/papers/axe/benchmark-llama3-groupwise.yml; src/brevitas_examples/papers/axe/benchmark.py; LICENSE
84
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d with its llm extras; the README (line 47) says the configs were tested with torch 2.6.0, transformers 4.57.6 and lighteval 0.13.0. Llama 3.2 1B Instruct weights from Hugging Face (meta-llama, gated, Llama 3.2 Community Licence; allowed with the gate noted). WikiText-2 raw test split (Salesforce wikitext on Hugging Face, CC BY-SA 3.0 / GFDL). lighteval zero-shot tasks ARC-Challenge, ARC-Easy, WinoGrande, HellaSwag and PIQA (public benchmarks on Hugging Face, each under its own terms); task versions as in the pinned lighteval release.
85
  - `src/brevitas_examples/papers/perq/README.md`: Brevitas block-rotation post-training quantization (PeRQ, ICML 2026) of Llama 3.2 1B Instruct at W4A4: WikiText2 perplexity per rotation block size, with and without permutation.
86
  Format: Markdown results table plus YAML configs
87
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
88
- Reference result: aggregate_metric_reference
89
- Reference source: papers/perq/README.md lines 69-70: Wiki2 PPL per block size, 1 decimal (No Permute 35.9 to 16.2, PeRQ 18.2 to 16.2)
90
  Check: Run one block size and permutation entry of benchmark-rotation_block_size.yml (brevitas_ptq_llm) at the pinned commit under python 3.12 and torch 2.6.0, and compare the WikiText2 PPL with the table cell within the tolerance below.
91
  Companions (not included): src/brevitas_examples/papers/perq/llama3-perq_star-int4.yml; src/brevitas_examples/papers/perq/llama3-perq_dag-int4.yml; src/brevitas_examples/papers/perq/benchmark-rotation_block_size.yml; src/brevitas_examples/papers/perq/benchmark.py; LICENSE
92
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d with its llm extras; python 3.12 and torch 2.6.0 per README line 65. Llama 3.2 1B Instruct weights from Hugging Face (meta-llama, gated, Llama 3.2 Community Licence; allowed with the gate noted). WikiText-2 raw test split (Salesforce wikitext on Hugging Face, CC BY-SA 3.0 / GFDL).
93
  - `src/brevitas_examples/papers/learned_round/README.md`: Brevitas learned-round post-training quantization of Llama 3.2 and Qwen 2.5 base models: WikiText2 perplexity and zero-shot average per transform and rounding method (W2g128 to MXFP4).
94
  Format: Markdown with HTML results tables plus YAML sweep configs
95
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
96
- Reference result: aggregate_metric_reference
97
- Reference source: papers/learned_round/README.md HTML tables from lines 76, 445, 821 and 1063: WikiText2 PPL and 0-shot normalized average accuracy (ARC-C, ARC-E, HellaSwag, PIQA, Winogrande; lines 28-35), 1 decimal; cells marked N/A are out-of-memory failures (line 64) and carry no answer
98
  Check: Run one stage 1 and stage 2 combination of one configuration for an in-scope model size (not the Qwen 2.5 3B columns) (the matching entry of the yaml named for that table, via benchmark.py and brevitas_ptq_llm) at the pinned commit with the versions of README line 6, and compare WikiText2 PPL and the 0-shot average with the table cell within the tolerance below; N/A cells are skipped.
99
  Companions (not included): src/brevitas_examples/papers/learned_round/learned_round_weight_only.yaml; src/brevitas_examples/papers/learned_round/learned_round_weight_only_spinquant.yaml; src/brevitas_examples/papers/learned_round/learned_round_weight_act.yaml; src/brevitas_examples/papers/learned_round/learned_round_weight_act_spinquant.yaml; src/brevitas_examples/papers/learned_round/learned_round_mxfp4.yaml; src/brevitas_examples/papers/learned_round/learned_round_mxfp4_spinquant.yaml; src/brevitas_examples/papers/learned_round/benchmark.py; LICENSE
100
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d with its llm extras; the README (line 6) says the yaml files work with brevitas 0.12.1, transformers 4.47.1 and lighteval 0.9.2. Llama 3.2 1B and 3B and Llama 3.1 8B base model weights (the tables label them Llama 3.2 with columns 1B, 3B, 8B) from Hugging Face (meta-llama, gated, Llama 3.2 and 3.1 Community Licences; allowed with the gate noted); Qwen 2.5 1.5B and 7B base weights (Apache-2.0 per their model cards); Qwen 2.5 3B base weights are under the Qwen Research License (non-commercial), so the 3B columns of the Qwen tables are out of scope for this row. WikiText-2 raw test split (Salesforce wikitext on Hugging Face, CC BY-SA 3.0 / GFDL). lighteval zero-shot tasks ARC-Challenge, ARC-Easy, WinoGrande, HellaSwag and PIQA (public benchmarks on Hugging Face, each under its own terms); task versions as in the pinned lighteval release.
 
8
 
9
  **Self-contained:** Whether each file needs other upstream files is not stated. 10 of 10 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Tools are named in the check written per file in the notes.
12
 
13
  **Licence (SPDX):** BSD-3-Clause
 
19
  - `src/brevitas_examples/imagenet_classification/ptq/RESULTS_TORCHVISION_BEST_CONFIGS.csv`: Brevitas post-training quantization of torchvision ImageNet models: best configuration per hardware feature set, with float and quantized top-1 (RESULTS_TORCHVISION_BEST_CONFIGS.csv).
20
  Format: CSV result file plus Python evaluation scripts and YAML config template
21
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
22
  Check: Pick one CSV row, run brevitas_ptq_imagenet_val (ptq_evaluate.py) at the pinned commit with that row's settings (model, target backend, scale factor type, weight, activation and bias bit widths, per-channel scaling, activation quant type, bias correction, graph equalization iterations and merge bias, GPTQ and act order, activation percentile, calibration size 1000 and batch size 64) on the ImageNet validation set, using torch 1.13.1 and brevitas at the version of the row; pass when the recomputed float and quantized top-1 match the printed values within the tolerance below. Rows made with other library versions are not comparable; the README warns results may vary across versions.
23
  Companions (not included): src/brevitas_examples/imagenet_classification/ptq/README.md; src/brevitas_examples/imagenet_classification/ptq/ptq_evaluate.py; src/brevitas_examples/imagenet_classification/ptq/ptq_common.py; src/brevitas_examples/imagenet_classification/ptq/ptq_imagenet_args.py; src/brevitas_examples/imagenet_classification/ptq/utils.py; src/brevitas_examples/imagenet_classification/ptq/config/default_template.yaml; src/brevitas_examples/imagenet_classification/ptq/benchmark/ptq_benchmark.py; src/brevitas_examples/imagenet_classification/ptq/benchmark/benchmark_template.yaml; src/brevitas_examples/imagenet_classification/utils.py; LICENSE
24
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (setup.py, requirements/requirements.txt) and torchvision pretrained weights for the 3 models in the CSV (mobilenet_v2, resnet18, vit_b_32 as named in column 1; BSD-3-Clause torchvision weights); the CSV was produced with torch 1.13.1 and brevitas 0.9.2.dev13+g96ddb35e (last two columns), so a recompute pins those versions. ImageNet-1k (ILSVRC2012) validation set, 50,000 images, gated, plus a calibration folder of 1,000 training images (ptq_imagenet_args.py line 48, --calibration-samples default 1000).
25
  - `src/brevitas_examples/imagenet_classification/ptq/RESULTS_IMGCLSMOB.csv`: Brevitas post-training quantization of the imgclsmob quant_mobilenet_v1 on ImageNet: float vs quantized top-1 (RESULTS_IMGCLSMOB.csv).
26
  Format: CSV result file plus Python evaluation scripts and an INI model config
27
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
28
  Check: Run brevitas_ptq_imagenet_val at the pinned commit for the imgclsmob quant_mobilenet_v1 with bias correction, calibration size 1000 and batch size 64, under torch 1.13.1 and the brevitas version of the row; pass when float and quantized top-1 match the printed values within the tolerances below.
29
  Companions (not included): src/brevitas_examples/imagenet_classification/ptq/README.md; src/brevitas_examples/imagenet_classification/ptq/ptq_evaluate.py; src/brevitas_examples/imagenet_classification/ptq/ptq_common.py; src/brevitas_examples/imagenet_classification/ptq/ptq_imagenet_args.py; src/brevitas_examples/imagenet_classification/ptq/utils.py; src/brevitas_examples/imagenet_classification/models/cfg/quant_mobilenet_v1.ini; src/brevitas_examples/imagenet_classification/models/README.md; src/brevitas_examples/imagenet_classification/utils.py; LICENSE
30
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d and the float MobileNet V1 weights the model config names (imgclsmob / pytorchcv, MIT per the source headers of models/mobilenetv1.py); the CSV was produced with torch 1.13.1 and brevitas 0.8.1.dev98, so a recompute pins those versions. ImageNet-1k (ILSVRC2012) validation set, 50,000 images, gated, plus a calibration folder of 1,000 training images (ptq_imagenet_args.py line 48, --calibration-samples default 1000).
31
  - `src/brevitas_examples/imagenet_classification/qat/README.md`: Brevitas quantization-aware trained ImageNet models (4-bit MobileNet V1 and ProxylessNAS Mobile14): published top-1 with released checkpoints.
32
  Format: Markdown results table plus Python evaluation script and INI model configs
33
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
34
  Check: For one model, run `brevitas_imagenet_val --imagenet-dir <imagenet> --model <cfg name> --pretrained` (qat/README.md line 24; the checkpoint is downloaded from the release) at the pinned commit and compare top-1 with the README value within the tolerance below.
35
  Companions (not included): src/brevitas_examples/imagenet_classification/qat/imagenet_val.py; src/brevitas_examples/imagenet_classification/models/cfg/quant_mobilenet_v1_4b.ini; src/brevitas_examples/imagenet_classification/models/cfg/quant_proxylessnas_mobile14_hadamard_4b.ini; src/brevitas_examples/imagenet_classification/models/cfg/quant_proxylessnas_mobile14_4b.ini; src/brevitas_examples/imagenet_classification/models/cfg/quant_proxylessnas_mobile14_4b5b.ini; src/brevitas_examples/imagenet_classification/models/README.md; src/brevitas_examples/imagenet_classification/utils.py; LICENSE
36
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (package brevitas_examples provides the model definitions models/mobilenetv1.py and models/proxylessnas.py, MIT, adapted from imgclsmob). the 4 quantized checkpoints released on GitHub (github.com/Xilinx/brevitas/releases, tags quant_mobilenet_v1_4b-r1 and quant_proxylessnas_mobile14_*-r0, files named in README lines 10-13 with hash suffixes); release assets under the repository licence. ImageNet-1k (ILSVRC2012) validation set, 50,000 images, gated.
37
  - `src/brevitas_examples/imagenet_classification/a2q/README.md`: Brevitas accumulator-aware quantization (A2Q, A2Q+) of ResNet18 on CIFAR10: published top-1 per target accumulator width with released checkpoints.
38
  Format: Markdown results table plus Python evaluation script
39
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
40
  Check: Run a2q_evaluate_models.py at the pinned commit for one model name of the table with the released checkpoint on the CIFAR-10 test set, and compare top-1 with the README value within the tolerance below.
41
  Companions (not included): src/brevitas_examples/imagenet_classification/a2q/a2q_evaluate_models.py; src/brevitas_examples/imagenet_classification/a2q/utils.py; LICENSE
42
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (package brevitas_examples provides a2q/resnet.py, MIT, adapted from kuangliu pytorch-cifar, and quant.py, ep_init.py). the 11 checkpoints released on GitHub (github.com/Xilinx/brevitas/releases tag a2q_cifar10_r1, file names with hash suffixes in README lines 16-28). CIFAR-10 test set (10,000 images, University of Toronto, no licence stated on the dataset page); downloaded by torchvision at run time.
43
  - `src/brevitas_examples/bnn_pynq/README.md`: Brevitas binarized and low-bit networks for FINN/PYNQ (TFC, SFC, LFC on MNIST; CNV and ResNet18 on CIFAR10): published top-1 with pretrained checkpoints.
44
  Format: Markdown results table plus Python evaluation scripts and INI model configs
45
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
46
  Check: For one model, run `BREVITAS_JIT=1 python bnn_pynq_train.py --evaluate --network <NAME> --pretrained` (README line 41) at the pinned commit on the test set named in the table, and compare top-1 with the README value within the tolerance below.
47
  Companions (not included): src/brevitas_examples/bnn_pynq/bnn_pynq_train.py; src/brevitas_examples/bnn_pynq/trainer.py; src/brevitas_examples/bnn_pynq/logger.py; src/brevitas_examples/bnn_pynq/cfg/tfc_1w1a.ini; src/brevitas_examples/bnn_pynq/cfg/tfc_1w2a.ini; src/brevitas_examples/bnn_pynq/cfg/tfc_2w2a.ini; src/brevitas_examples/bnn_pynq/cfg/sfc_1w1a.ini; src/brevitas_examples/bnn_pynq/cfg/sfc_1w2a.ini; src/brevitas_examples/bnn_pynq/cfg/sfc_2w2a.ini; src/brevitas_examples/bnn_pynq/cfg/lfc_1w1a.ini; src/brevitas_examples/bnn_pynq/cfg/lfc_1w2a.ini; src/brevitas_examples/bnn_pynq/cfg/cnv_1w1a.ini; src/brevitas_examples/bnn_pynq/cfg/cnv_1w2a.ini; src/brevitas_examples/bnn_pynq/cfg/cnv_2w2a.ini; src/brevitas_examples/bnn_pynq/cfg/resnet18_4w4a.ini; LICENSE
48
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (package brevitas_examples provides models/*.py). pretrained checkpoints whose URLs are in the cfg/*.ini files (GitHub releases of Xilinx/brevitas). MNIST test set (10,000 images; Yann LeCun and Corinna Cortes; licence statements vary by mirror) for the 8 FC models and CIFAR-10 test set (10,000 images; no licence stated) for the 4 CIFAR models; downloaded by torchvision at run time.
49
  - `src/brevitas_examples/speech_to_text/README.md`: Brevitas quantized QuartzNet 15x5 speech recognition (8-bit and 4-bit): published WER on LibriSpeech dev-other with released checkpoints.
50
  Format: Markdown results table plus Python evaluation script, INI and YAML configs
51
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
52
  Check: Prepare LibriSpeech dev-other with brevitas_quartznet_preprocess, run `brevitas_quartznet_val --data-json <json> --model <cfg name> --pretrained` (README line 33) at the pinned commit, and compare the WER with the README value within the tolerance below.
53
  Companions (not included): src/brevitas_examples/speech_to_text/quartznet_val.py; src/brevitas_examples/speech_to_text/cfg/quant_quartznet_pertensorscaling_8b.ini; src/brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_8b.ini; src/brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_4b.ini; src/brevitas_examples/speech_to_text/cfg/topology/quartznet15x5.yaml; LICENSE
54
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d (package brevitas_examples provides speech_to_text/quartznet/, adapted from NVIDIA NeMo r0.9, Apache-2.0 headers). encoder and decoder checkpoints released on GitHub (tags quant_quartznet_8b-r0 and quant_quartznet_4b-r0, README lines 10-12). LibriSpeech dev-other (openslr.org SLR12, CC BY 4.0), prepared by brevitas_quartznet_preprocess (get_librispeech_data.py).
55
  - `src/brevitas_examples/papers/gguf/README.md`: Brevitas GGUF export validation: in-process perplexity vs llama.cpp llama-perplexity for 13 GGUF recipes on three Llama 3 models, with the Qronos Q2_K configs.
56
  Format: Markdown results tables plus YAML configs and Python recipe plugins
57
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
58
  Check: Export one recipe of one model to GGUF with brevitas at the pinned commit (fields custom_quantizer and export_target as in the README config section, lines 90-130, field table lines 96-98; the Qronos rows use the provided llama3-*-q2_k.yml), build wiki.test.raw as README lines 140-151 do, run `llama-perplexity -m <model.gguf> -f wiki.test.raw -c 2048 -ngl 0` (line 157, 141 chunks), and compare the llama.cpp PPL with the table value within the half-unit tolerance; the Brevitas in-process PPL is compared to the llama.cpp PPL of the same row within the published harness gap.
59
  Companions (not included): src/brevitas_examples/papers/gguf/llama3-1b-q2_k.yml; src/brevitas_examples/papers/gguf/llama3-3b-q2_k.yml; src/brevitas_examples/papers/gguf/llama3-8b-q2_k.yml; src/brevitas_examples/papers/gguf/recipes/Llama-3.2-1B.py; src/brevitas_examples/papers/gguf/recipes/Llama-3.2-3B.py; src/brevitas_examples/papers/gguf/recipes/Llama-3.1-8B.py; src/brevitas_examples/papers/gguf/recipes/common.py; LICENSE
60
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d with its llm extras (entry point brevitas_ptq_llm). llama.cpp llama-perplexity (MIT) at a pinned release. Llama 3.2 1B Instruct, Llama 3.2 3B Instruct and Llama 3.1 8B Instruct weights from Hugging Face (meta-llama, gated, Llama 3.2 and Llama 3.1 Community Licence; allowed with the gate noted). WikiText-2 raw test split (Salesforce wikitext on Hugging Face, CC BY-SA 3.0 / GFDL).
61
  - `src/brevitas_examples/papers/axe/README.md`: Brevitas accumulator-aware post-training quantization of LLMs (AXE, TMLR 2025) on Llama 3.2 1B Instruct: WikiText2 perplexity and five zero-shot accuracies per configuration.
62
  Format: Markdown results tables plus YAML sweep configs
63
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
64
  Check: Run one configuration of the sweep (the matching entry of benchmark-llama3-channelwise.yml or benchmark-llama3-groupwise.yml, via brevitas_ptq_llm) at the pinned commit with the pinned torch, transformers and lighteval, and compare Wiki2 PPL and each zero-shot accuracy with the table row within the tolerances below.
65
  Companions (not included): src/brevitas_examples/papers/axe/benchmark-llama3-channelwise.yml; src/brevitas_examples/papers/axe/benchmark-llama3-groupwise.yml; src/brevitas_examples/papers/axe/benchmark.py; LICENSE
66
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d with its llm extras; the README (line 47) says the configs were tested with torch 2.6.0, transformers 4.57.6 and lighteval 0.13.0. Llama 3.2 1B Instruct weights from Hugging Face (meta-llama, gated, Llama 3.2 Community Licence; allowed with the gate noted). WikiText-2 raw test split (Salesforce wikitext on Hugging Face, CC BY-SA 3.0 / GFDL). lighteval zero-shot tasks ARC-Challenge, ARC-Easy, WinoGrande, HellaSwag and PIQA (public benchmarks on Hugging Face, each under its own terms); task versions as in the pinned lighteval release.
67
  - `src/brevitas_examples/papers/perq/README.md`: Brevitas block-rotation post-training quantization (PeRQ, ICML 2026) of Llama 3.2 1B Instruct at W4A4: WikiText2 perplexity per rotation block size, with and without permutation.
68
  Format: Markdown results table plus YAML configs
69
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
70
  Check: Run one block size and permutation entry of benchmark-rotation_block_size.yml (brevitas_ptq_llm) at the pinned commit under python 3.12 and torch 2.6.0, and compare the WikiText2 PPL with the table cell within the tolerance below.
71
  Companions (not included): src/brevitas_examples/papers/perq/llama3-perq_star-int4.yml; src/brevitas_examples/papers/perq/llama3-perq_dag-int4.yml; src/brevitas_examples/papers/perq/benchmark-rotation_block_size.yml; src/brevitas_examples/papers/perq/benchmark.py; LICENSE
72
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d with its llm extras; python 3.12 and torch 2.6.0 per README line 65. Llama 3.2 1B Instruct weights from Hugging Face (meta-llama, gated, Llama 3.2 Community Licence; allowed with the gate noted). WikiText-2 raw test split (Salesforce wikitext on Hugging Face, CC BY-SA 3.0 / GFDL).
73
  - `src/brevitas_examples/papers/learned_round/README.md`: Brevitas learned-round post-training quantization of Llama 3.2 and Qwen 2.5 base models: WikiText2 perplexity and zero-shot average per transform and rounding method (W2g128 to MXFP4).
74
  Format: Markdown with HTML results tables plus YAML sweep configs
75
  Origin: published result tables of a quantization library, recorded by its maintainers from runs of the example scripts and configs in the same directory
 
 
76
  Check: Run one stage 1 and stage 2 combination of one configuration for an in-scope model size (not the Qwen 2.5 3B columns) (the matching entry of the yaml named for that table, via benchmark.py and brevitas_ptq_llm) at the pinned commit with the versions of README line 6, and compare WikiText2 PPL and the 0-shot average with the table cell within the tolerance below; N/A cells are skipped.
77
  Companions (not included): src/brevitas_examples/papers/learned_round/learned_round_weight_only.yaml; src/brevitas_examples/papers/learned_round/learned_round_weight_only_spinquant.yaml; src/brevitas_examples/papers/learned_round/learned_round_weight_act.yaml; src/brevitas_examples/papers/learned_round/learned_round_weight_act_spinquant.yaml; src/brevitas_examples/papers/learned_round/learned_round_mxfp4.yaml; src/brevitas_examples/papers/learned_round/learned_round_mxfp4_spinquant.yaml; src/brevitas_examples/papers/learned_round/benchmark.py; LICENSE
78
  Dependencies: brevitas installed from commit fda71bcb5b45bda7ee4ec1304b51e5c3be45610d with its llm extras; the README (line 6) says the yaml files work with brevitas 0.12.1, transformers 4.47.1 and lighteval 0.9.2. Llama 3.2 1B and 3B and Llama 3.1 8B base model weights (the tables label them Llama 3.2 with columns 1B, 3B, 8B) from Hugging Face (meta-llama, gated, Llama 3.2 and 3.1 Community Licences; allowed with the gate noted); Qwen 2.5 1.5B and 7B base weights (Apache-2.0 per their model cards); Qwen 2.5 3B base weights are under the Qwen Research License (non-commercial), so the 3B columns of the Qwen tables are out of scope for this row. WikiText-2 raw test split (Salesforce wikitext on Hugging Face, CC BY-SA 3.0 / GFDL). lighteval zero-shot tasks ARC-Challenge, ARC-Easy, WinoGrande, HellaSwag and PIQA (public benchmarks on Hugging Face, each under its own terms); task versions as in the pinned lighteval release.
campaign_files/2026-10-op-rule-calculations/xbrl-us-dqc-rules-tests/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Each file is marked self-contained upstream. 7 of 7 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_validation_findings. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Tools are named in the check written per file in the notes.
14
 
15
  **Licence (SPDX):** LicenseRef-XBRLUS-DQC-License
@@ -21,49 +19,35 @@ Files:
21
  - `tests/input/testfiles/2025/TestCo-2025-NonNeg-WithNew.xml`: XBRL US Data Quality Committee rule DQC.US.0015 (Negative Values): XBRL instance with the expected validation log from the official rule test suite.
22
  Format: XBRL 2.1 instance (XML); expected output is an Arelle XML log; rule in XULE text
23
  Origin: synthetic US GAAP test company instance built by XBRL US to exercise the non-negative element list; 6,471 DQC.US.0015 findings expected; expected log is the upstream regression output the DQC project checks its rules against; 031 full check: rule and companions repinned to 3b8ca1f (27.0.7) to match the log version
24
- Reference result: expected_validation_findings
25
- Reference source: tests/output/DQC.US.0015_testco_us-2025.xml: Arelle plus XULE validation log produced by XBRL US; 6471 entries with code DQC.US.0015* (info and preferredLabelMissing entries are processor noise and are filtered out); rule text, constants and lists pinned at commit 3b8ca1f (ruleVersion 27.0.7, same as the expected log); HEAD 16eb506 (30.0.5) changed DQC_0015.xule, the md, ListOfElements.xlsx, constant files and lib, so it is not used
26
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0015/DQC_0015.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0015 entries of the expected log, keyed on concept plus context; message wording is not scored
27
  Companions (not included): tests/output/DQC.US.0015_testco_us-2025.xml; dqc_us_rules/source/us/2025/DQC_0015.xule; docs/DQC_US_0015/DQC_0015.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; docs/DQC_US_0015/DQC_0015_ListOfElements.xlsx; docs/DQC_US_0015/DQC_0015_MemberExclusions.xlsx; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_usgaap_2025_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_srt_2025_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_dei_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_ebp_2025_concepts.csv
28
  - `tests/input/testfiles/2025/TestCo-2025-NonNeg-IFRS.xml`: XBRL US Data Quality Committee rule DQC.IFRS.0080 (Negative Values (IFRS)): XBRL instance with the expected validation log from the official rule test suite.
29
  Format: XBRL 2.1 instance (XML); expected output is an Arelle XML log; rule in XULE text
30
  Origin: synthetic IFRS test company instance; 1,554 DQC.IFRS.0080 findings expected; expected log is the upstream regression output the DQC project checks its rules against; 031 full check: rule and companions repinned to 2cf4a8c (28.0.1) to match the log version
31
- Reference result: expected_validation_findings
32
- Reference source: tests/output/DQC.IFRS.0080_testco_ifrs-2025.xml: Arelle plus XULE validation log produced by XBRL US; 1554 entries with code DQC.IFRS.0080* (info and preferredLabelMissing entries are processor noise and are filtered out); rule text, constants and lists pinned at commit 2cf4a8c (ruleVersion 28.0.1, same as the expected log); HEAD 16eb506 changed DQC_IFRS_0080.xule, the md, constant-IFRS.xule and lib, so it is not used
33
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_IFRS_0080/DQC_0080.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.IFRS.0080 entries of the expected log, keyed on concept plus context; message wording is not scored
34
  Companions (not included): tests/output/DQC.IFRS.0080_testco_ifrs-2025.xml; dqc_us_rules/source/ifrs/2025/DQC_IFRS_0080.xule; docs/DQC_IFRS_0080/DQC_0080.md; License.md; tests/readme.md; dqc_us_rules/source/ifrs/2025/constant-IFRS.xule; dqc_us_rules/source/ifrs/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; docs/DQC_IFRS_0080/DQC_0080_ListOfElements.xlsx; docs/DQC_IFRS_0080/DQC_0080_MemberExclusions.xlsx; dqc_us_rules/resources/Effective_Dates/ifrs_effective_dates.csv; dqc_us_rules/resources/DQC_IFRS_0080/dqc_0080_ifrs_2025_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_srt_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_dei_concepts.csv
35
  - `tests/input/testfiles/2024-roll-2025/OGS_instance_2024_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0001.51 (Axis with Inappropriate Members): XBRL instance with the expected validation log from the official rule test suite.
36
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
37
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (ONE Gas, Inc., CIK 0001587732, 10-Q for period 2024-09-30, accession 0001587732-24-000060, filed 2024-11-05, https://www.sec.gov/Archives/edgar/data/1587732/000158773224000060/ogs-20240930.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API
38
- Reference result: expected_validation_findings
39
- Reference source: tests/output/rollDQC.US.0001.51_OGS-us-2024.xml: Arelle plus XULE validation log produced by XBRL US; 2 entries with code DQC.US.0001.51* (info and preferredLabelMissing entries are processor noise and are filtered out); rule text pinned at 30da4d5: DQC_0001.xule and its md are the same blobs at 555d54e (29.0.2), 30da4d5 and HEAD; constant.xule and constant-us-gaap.xule changed after 29.0.2 (30.0.0 and 30.0.5), so the constant files are pinned at 30da4d5, where among rule files only version.xule differs from 555d54e
40
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0001/DQC_0001.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0001.51 entries of the expected log, keyed on concept plus context; message wording is not scored
41
  Companions (not included): tests/input/testfiles/2024-roll-2025/OGS_taxonomy_2024_2025.zip; tests/output/rollDQC.US.0001.51_OGS-us-2024.xml; dqc_us_rules/source/us/2025/DQC_0001.xule; docs/DQC_US_0001/DQC_0001.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv; dqc_us_rules/source/us/2025/base-taxonomy-2025.xule; dqc_us_rules/resources/Taxonomies/us-gaap-cal-2025-all.xsd
42
  - `tests/input/testfiles/2024-roll-2025/ENRT_instance_2024_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0004.16 (Element Values Are Equal): XBRL instance with the expected validation log from the official rule test suite.
43
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
44
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (Enertopia Corp, CIK 0001346022, 10-Q for period 2024-11-30, accession 0001062993-25-000717, filed 2025-01-13, https://www.sec.gov/Archives/edgar/data/1346022/000106299325000717/form10q.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API
45
- Reference result: expected_validation_findings
46
- Reference source: tests/output/rollDQC.US.0004.16_ENRT-us-2024.xml: Arelle plus XULE validation log produced by XBRL US; 1 entries with code DQC.US.0004.16* (info and preferredLabelMissing entries are processor noise and are filtered out); rule text pinned at 30da4d5: DQC_0004.xule and its md are the same blobs at 555d54e (29.0.2), 30da4d5 and HEAD; constant.xule and constant-us-gaap.xule changed after 29.0.2 (30.0.0 and 30.0.5), so the constant files are pinned at 30da4d5, where among rule files only version.xule differs from 555d54e
47
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0004/DQC_0004.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0004.16 entries of the expected log, keyed on concept plus context; message wording is not scored
48
  Companions (not included): tests/input/testfiles/2024-roll-2025/ENRT_taxonomy_2024_2025.zip; tests/output/rollDQC.US.0004.16_ENRT-us-2024.xml; dqc_us_rules/source/us/2025/DQC_0004.xule; docs/DQC_US_0004/DQC_0004.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv
49
  - `tests/input/testfiles/2023-roll-2025/CARE_instance_2023_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0013 (Negative Values With Dependence): XBRL instance with the expected validation log from the official rule test suite.
50
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
51
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (Carter Bankshares, Inc., CIK 0001829576, 10-Q for period 2023-06-30, accession 0001829576-23-000043, filed 2023-08-03, https://www.sec.gov/Archives/edgar/data/1829576/000182957623000043/care-20230630.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API
52
- Reference result: expected_validation_findings
53
- Reference source: tests/output/rollDQC.US.0013_CARE-us-2023.xml: Arelle plus XULE validation log produced by XBRL US; 5 entries with code DQC.US.0013* (info and preferredLabelMissing entries are processor noise and are filtered out); rule text pinned at 30da4d5: DQC_0013.xule and its md are the same blobs at 555d54e (29.0.2), 30da4d5 and HEAD; constant.xule and constant-us-gaap.xule changed after 29.0.2 (30.0.0 and 30.0.5), so the constant files are pinned at 30da4d5, where among rule files only version.xule differs from 555d54e
54
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0013/DQC_0013.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0013 entries of the expected log, keyed on concept plus context; message wording is not scored
55
  Companions (not included): tests/input/testfiles/2023-roll-2025/CARE_taxonomy_2023_2025.zip; tests/output/rollDQC.US.0013_CARE-us-2023.xml; dqc_us_rules/source/us/2025/DQC_0013.xule; docs/DQC_US_0013/DQC_0013.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv
56
  - `tests/input/testfiles/2024-roll-2025/CEI_instance_2024_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0046.9279 (Inappropriate Calculation Descendent): XBRL instance with the expected validation log from the official rule test suite.
57
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
58
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (Camber Energy, Inc. (now Viking Global Technologies), CIK 0001309082, 10-Q/A for period 2023-09-30, accession 0001477932-24-005253, filed 2024-08-26, https://www.sec.gov/Archives/edgar/data/1309082/000147793224005253/cei_10q.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API; the instance carries AmendmentFlag true; the original 10-Q is 0001477932-23-008345 filed 2023-11-14
59
- Reference result: expected_validation_findings
60
- Reference source: tests/output/rollDQC.US.0046.9279_CEI-us-2024.xml: Arelle plus XULE validation log produced by XBRL US; 1 entries with code DQC.US.0046.9279* (info and preferredLabelMissing entries are processor noise and are filtered out); rule text pinned at 30da4d5: DQC_0046.xule and its md are the same blobs at 555d54e (29.0.2), 30da4d5 and HEAD; constant.xule and constant-us-gaap.xule changed after 29.0.2 (30.0.0 and 30.0.5), so the constant files are pinned at 30da4d5, where among rule files only version.xule differs from 555d54e
61
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0046/DQC_0046.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0046.9279 entries of the expected log, keyed on concept plus context; message wording is not scored
62
  Companions (not included): tests/input/testfiles/2024-roll-2025/CEI_taxonomy_2024_2025.zip; tests/output/rollDQC.US.0046.9279_CEI-us-2024.xml; dqc_us_rules/source/us/2025/DQC_0046.xule; docs/DQC_US_0046/DQC_0046.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv
63
  - `tests/input/testfiles/2024-roll-2025/SPR_instance_2024_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0055.7493 (Required Member on An Axis): XBRL instance with the expected validation log from the official rule test suite.
64
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
65
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (Spirit AeroSystems Holdings, Inc., CIK 0001364885, 10-K for period 2024-12-31, accession 0001628280-25-009088, filed 2025-02-28, https://www.sec.gov/Archives/edgar/data/1364885/000162828025009088/spr-20241231.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API
66
- Reference result: expected_validation_findings
67
- Reference source: tests/output/rollDQC.US.0055.7493_SPR-us-2024.xml: Arelle plus XULE validation log produced by XBRL US; 1 entries with code DQC.US.0055.7493* (info and preferredLabelMissing entries are processor noise and are filtered out); rule text pinned at 30da4d5: DQC_0055.xule and its md are the same blobs at 555d54e (29.0.2), 30da4d5 and HEAD; constant.xule and constant-us-gaap.xule changed after 29.0.2 (30.0.0 and 30.0.5), so the constant files are pinned at 30da4d5, where among rule files only version.xule differs from 555d54e
68
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0055/DQC_0055.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0055.7493 entries of the expected log, keyed on concept plus context; message wording is not scored
69
  Companions (not included): tests/input/testfiles/2024-roll-2025/SPR_taxonomy_2024_2025.zip; tests/output/rollDQC.US.0055.7493_SPR-us-2024.xml; dqc_us_rules/source/us/2025/DQC_0055.xule; docs/DQC_US_0055/DQC_0055.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv
 
8
 
9
  **Self-contained:** Each file is marked self-contained upstream. 7 of 7 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Tools are named in the check written per file in the notes.
12
 
13
  **Licence (SPDX):** LicenseRef-XBRLUS-DQC-License
 
19
  - `tests/input/testfiles/2025/TestCo-2025-NonNeg-WithNew.xml`: XBRL US Data Quality Committee rule DQC.US.0015 (Negative Values): XBRL instance with the expected validation log from the official rule test suite.
20
  Format: XBRL 2.1 instance (XML); expected output is an Arelle XML log; rule in XULE text
21
  Origin: synthetic US GAAP test company instance built by XBRL US to exercise the non-negative element list; 6,471 DQC.US.0015 findings expected; expected log is the upstream regression output the DQC project checks its rules against; 031 full check: rule and companions repinned to 3b8ca1f (27.0.7) to match the log version
 
 
22
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0015/DQC_0015.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0015 entries of the expected log, keyed on concept plus context; message wording is not scored
23
  Companions (not included): tests/output/DQC.US.0015_testco_us-2025.xml; dqc_us_rules/source/us/2025/DQC_0015.xule; docs/DQC_US_0015/DQC_0015.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; docs/DQC_US_0015/DQC_0015_ListOfElements.xlsx; docs/DQC_US_0015/DQC_0015_MemberExclusions.xlsx; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_usgaap_2025_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_srt_2025_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_dei_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_ebp_2025_concepts.csv
24
  - `tests/input/testfiles/2025/TestCo-2025-NonNeg-IFRS.xml`: XBRL US Data Quality Committee rule DQC.IFRS.0080 (Negative Values (IFRS)): XBRL instance with the expected validation log from the official rule test suite.
25
  Format: XBRL 2.1 instance (XML); expected output is an Arelle XML log; rule in XULE text
26
  Origin: synthetic IFRS test company instance; 1,554 DQC.IFRS.0080 findings expected; expected log is the upstream regression output the DQC project checks its rules against; 031 full check: rule and companions repinned to 2cf4a8c (28.0.1) to match the log version
 
 
27
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_IFRS_0080/DQC_0080.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.IFRS.0080 entries of the expected log, keyed on concept plus context; message wording is not scored
28
  Companions (not included): tests/output/DQC.IFRS.0080_testco_ifrs-2025.xml; dqc_us_rules/source/ifrs/2025/DQC_IFRS_0080.xule; docs/DQC_IFRS_0080/DQC_0080.md; License.md; tests/readme.md; dqc_us_rules/source/ifrs/2025/constant-IFRS.xule; dqc_us_rules/source/ifrs/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; docs/DQC_IFRS_0080/DQC_0080_ListOfElements.xlsx; docs/DQC_IFRS_0080/DQC_0080_MemberExclusions.xlsx; dqc_us_rules/resources/Effective_Dates/ifrs_effective_dates.csv; dqc_us_rules/resources/DQC_IFRS_0080/dqc_0080_ifrs_2025_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_srt_concepts.csv; dqc_us_rules/resources/DQC_US_0015/dqc_15_dei_concepts.csv
29
  - `tests/input/testfiles/2024-roll-2025/OGS_instance_2024_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0001.51 (Axis with Inappropriate Members): XBRL instance with the expected validation log from the official rule test suite.
30
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
31
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (ONE Gas, Inc., CIK 0001587732, 10-Q for period 2024-09-30, accession 0001587732-24-000060, filed 2024-11-05, https://www.sec.gov/Archives/edgar/data/1587732/000158773224000060/ogs-20240930.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API
 
 
32
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0001/DQC_0001.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0001.51 entries of the expected log, keyed on concept plus context; message wording is not scored
33
  Companions (not included): tests/input/testfiles/2024-roll-2025/OGS_taxonomy_2024_2025.zip; tests/output/rollDQC.US.0001.51_OGS-us-2024.xml; dqc_us_rules/source/us/2025/DQC_0001.xule; docs/DQC_US_0001/DQC_0001.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv; dqc_us_rules/source/us/2025/base-taxonomy-2025.xule; dqc_us_rules/resources/Taxonomies/us-gaap-cal-2025-all.xsd
34
  - `tests/input/testfiles/2024-roll-2025/ENRT_instance_2024_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0004.16 (Element Values Are Equal): XBRL instance with the expected validation log from the official rule test suite.
35
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
36
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (Enertopia Corp, CIK 0001346022, 10-Q for period 2024-11-30, accession 0001062993-25-000717, filed 2025-01-13, https://www.sec.gov/Archives/edgar/data/1346022/000106299325000717/form10q.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API
 
 
37
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0004/DQC_0004.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0004.16 entries of the expected log, keyed on concept plus context; message wording is not scored
38
  Companions (not included): tests/input/testfiles/2024-roll-2025/ENRT_taxonomy_2024_2025.zip; tests/output/rollDQC.US.0004.16_ENRT-us-2024.xml; dqc_us_rules/source/us/2025/DQC_0004.xule; docs/DQC_US_0004/DQC_0004.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv
39
  - `tests/input/testfiles/2023-roll-2025/CARE_instance_2023_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0013 (Negative Values With Dependence): XBRL instance with the expected validation log from the official rule test suite.
40
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
41
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (Carter Bankshares, Inc., CIK 0001829576, 10-Q for period 2023-06-30, accession 0001829576-23-000043, filed 2023-08-03, https://www.sec.gov/Archives/edgar/data/1829576/000182957623000043/care-20230630.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API
 
 
42
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0013/DQC_0013.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0013 entries of the expected log, keyed on concept plus context; message wording is not scored
43
  Companions (not included): tests/input/testfiles/2023-roll-2025/CARE_taxonomy_2023_2025.zip; tests/output/rollDQC.US.0013_CARE-us-2023.xml; dqc_us_rules/source/us/2025/DQC_0013.xule; docs/DQC_US_0013/DQC_0013.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv
44
  - `tests/input/testfiles/2024-roll-2025/CEI_instance_2024_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0046.9279 (Inappropriate Calculation Descendent): XBRL instance with the expected validation log from the official rule test suite.
45
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
46
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (Camber Energy, Inc. (now Viking Global Technologies), CIK 0001309082, 10-Q/A for period 2023-09-30, accession 0001477932-24-005253, filed 2024-08-26, https://www.sec.gov/Archives/edgar/data/1309082/000147793224005253/cei_10q.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API; the instance carries AmendmentFlag true; the original 10-Q is 0001477932-23-008345 filed 2023-11-14
 
 
47
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0046/DQC_0046.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0046.9279 entries of the expected log, keyed on concept plus context; message wording is not scored
48
  Companions (not included): tests/input/testfiles/2024-roll-2025/CEI_taxonomy_2024_2025.zip; tests/output/rollDQC.US.0046.9279_CEI-us-2024.xml; dqc_us_rules/source/us/2025/DQC_0046.xule; docs/DQC_US_0046/DQC_0046.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv
49
  - `tests/input/testfiles/2024-roll-2025/SPR_instance_2024_2025.xml`: XBRL US Data Quality Committee rule DQC.US.0055.7493 (Required Member on An Axis): XBRL instance with the expected validation log from the official rule test suite.
50
  Format: XBRL 2.1 instance (XML) plus extension taxonomy zip; expected output is an Arelle XML log; rule in XULE text
51
  Origin: test fixture made by XBRL US: the inline XBRL of a real SEC EDGAR filing (Spirit AeroSystems Holdings, Inc., CIK 0001364885, 10-K for period 2024-12-31, accession 0001628280-25-009088, filed 2025-02-28, https://www.sec.gov/Archives/edgar/data/1364885/000162828025009088/spr-20241231.htm) re-encoded as an XBRL instance and rolled forward to the 2025 taxonomy (instance comment: Created by Xince), added to the DQC test set at commit 30da4d5 (v29.0.3) with its expected log (ruleVersion 29.0.2); original traced via the dei facts in the instance and the EDGAR submissions API
 
 
52
  Check: give the agent the instance (and taxonomy zip), the rule documentation docs/DQC_US_0055/DQC_0055.md and the XULE rule; it lists every fact that breaks the rule as (rule code, concept, context or period, value); score set F1 or exact match against the DQC.US.0055.7493 entries of the expected log, keyed on concept plus context; message wording is not scored
53
  Companions (not included): tests/input/testfiles/2024-roll-2025/SPR_taxonomy_2024_2025.zip; tests/output/rollDQC.US.0055.7493_SPR-us-2024.xml; dqc_us_rules/source/us/2025/DQC_0055.xule; docs/DQC_US_0055/DQC_0055.md; License.md; tests/readme.md; dqc_us_rules/source/us/2025/constant.xule; dqc_us_rules/source/us/2025/constant-us-gaap.xule; dqc_us_rules/source/us/2025/resources.xule; dqc_us_rules/source/lib/functions.xule; dqc_us_rules/source/lib/version.xule; dqc_us_rules/resources/Effective_Dates/effective_dates.csv; dqc_us_rules/resources/rule-form-lookup/rule-form-lookup.csv
campaign_files/2026-10-s-security-detection-2/welcome-ctf-2025-public-challenges/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 5 of 5 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: flag_string. Where the reference result comes from and how a solution is checked is written per unit in the notes.
12
-
13
  **Tools to open and check:** Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance.
14
 
15
  **Licence (SPDX):** NOASSERTION
@@ -21,40 +19,30 @@ Units:
21
  - `forensics/baby_wireshark/dist/login_capture.pcap`: Welcome CTF 2025 forensics/Baby Wireshark: from a pcap of a brute-force against a Flask login form, find the one username and password pair that succeeded.
22
  Format: one network capture (pcap, 1,898,180 B) of HTTP POSTs to a Flask login endpoint; many wrong credential attempts and one success
23
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
24
- Reference result: flag_string
25
- Reference source: README.md line 14 and challenge.yml line 21 give the flag grey{hunterxhunter_abc123}. README says the correct pair is the one accepted by the server. Partly verified offline: the capture holds 1,172 login POSTs; the single server response with Content-Length 20 (a short success body) follows the POST carrying username hunterxhunter, matching the flag.
26
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
27
  Companions (not included): challenge.yml; README.md
28
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
29
  - `forensics/shy_zipper/distrib/shy_zipper.zip`: Welcome CTF 2025 forensics/Shy Zipper: a zip with bytes appended after its End Of Central Directory record; recover the flag hidden outside the zip structure.
30
  Format: one zip archive (3,418 B) with extra bytes appended after the EOCD record (data living outside the central directory)
31
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
32
- Reference result: flag_string
33
- Reference source: README.md line 12 and challenge.yml line 14 give grey{s0m3_Th1Ng5_4r3_b3tT3r_L3fT_uNz1Pp3d}; solve/README.md gives the path (read the bytes after the EOCD, base64 decode). Verified offline: the trailing bytes after the EOCD base64-decode to exactly this flag
34
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
35
  Companions (not included): challenge.yml; README.md; solve/README.md
36
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
37
  - `forensics/cosmic_bit_flip/dist/flag.png`: Welcome CTF 2025 forensics/Cosmic Bit Flip: a PNG that renders as static; recover the flag drawn in the image (an opaque RGB with no usable alpha).
38
  Format: one PNG image (856,267 B, RGB 1029x242, no alpha channel) that appears as noise until the pixels are interpreted
39
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
40
- Reference result: flag_string
41
- Reference source: README.md line 12 and challenge.yml line 17 give flag{b1t_fl1p_m4d3_my_fl4g_tr4nsp4r3n7}. The flag is read by a visual check of the recovered image.
42
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
43
  Companions (not included): challenge.yml; README.md
44
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
45
  - `forensics/scrambled_pawtrait/` (2 files): Welcome CTF 2025 forensics/Scrambled Pawtrait: two same-size images are the XOR halves of one picture; XOR them per pixel and channel to read the flag.
46
  Format: two PNG images of equal size (512x512 RGB each) that recombine by a per-pixel per-channel bitwise XOR
47
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
48
- Reference result: flag_string
49
- Reference source: README.md line 11 and challenge.yml line 14 give grey{purrf3ct_x0r}; solve/README.md gives the path (XOR the two images).
50
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
51
  Companions (not included): challenge.yml; README.md; solve/README.md
52
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
53
  - `forensics/solved_hangman/dist/flag.png`: Welcome CTF 2025 forensics/Solved Hangman: a file saved with a corrupted (masqueraded) PNG header; repair the header and read the flag from the picture.
54
  Format: one image file (8,850 B) whose magic bytes were overwritten to a bogus GREY\r\n\x1a\n header instead of the PNG signature
55
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
56
- Reference result: flag_string
57
- Reference source: README.md line 8 and challenge.yml line 13 give grey{gheiii}. Verified offline that the file header is corrupted (bytes GREY\r\n\x1a\n in place of the PNG signature).
58
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
59
  Companions (not included): challenge.yml; README.md
60
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
 
8
 
9
  **Self-contained:** Not self-contained: each unit is used together with other files from the same upstream. 5 of 5 units name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per unit in the notes. Dependencies named upstream are listed per unit in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance.
12
 
13
  **Licence (SPDX):** NOASSERTION
 
19
  - `forensics/baby_wireshark/dist/login_capture.pcap`: Welcome CTF 2025 forensics/Baby Wireshark: from a pcap of a brute-force against a Flask login form, find the one username and password pair that succeeded.
20
  Format: one network capture (pcap, 1,898,180 B) of HTTP POSTs to a Flask login endpoint; many wrong credential attempts and one success
21
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
 
 
22
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
23
  Companions (not included): challenge.yml; README.md
24
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
25
  - `forensics/shy_zipper/distrib/shy_zipper.zip`: Welcome CTF 2025 forensics/Shy Zipper: a zip with bytes appended after its End Of Central Directory record; recover the flag hidden outside the zip structure.
26
  Format: one zip archive (3,418 B) with extra bytes appended after the EOCD record (data living outside the central directory)
27
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
 
 
28
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
29
  Companions (not included): challenge.yml; README.md; solve/README.md
30
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
31
  - `forensics/cosmic_bit_flip/dist/flag.png`: Welcome CTF 2025 forensics/Cosmic Bit Flip: a PNG that renders as static; recover the flag drawn in the image (an opaque RGB with no usable alpha).
32
  Format: one PNG image (856,267 B, RGB 1029x242, no alpha channel) that appears as noise until the pixels are interpreted
33
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
 
 
34
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
35
  Companions (not included): challenge.yml; README.md
36
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
37
  - `forensics/scrambled_pawtrait/` (2 files): Welcome CTF 2025 forensics/Scrambled Pawtrait: two same-size images are the XOR halves of one picture; XOR them per pixel and channel to read the flag.
38
  Format: two PNG images of equal size (512x512 RGB each) that recombine by a per-pixel per-channel bitwise XOR
39
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
 
 
40
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
41
  Companions (not included): challenge.yml; README.md; solve/README.md
42
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
43
  - `forensics/solved_hangman/dist/flag.png`: Welcome CTF 2025 forensics/Solved Hangman: a file saved with a corrupted (masqueraded) PNG header; repair the header and read the flag from the picture.
44
  Format: one image file (8,850 B) whose magic bytes were overwritten to a bogus GREY\r\n\x1a\n header instead of the PNG signature
45
  Origin: official competition instance built by the organisers for a public CTF and released with its flag and solution (fixture of a competition, not recorded production telemetry)
 
 
46
  Check: Exact string match of the submitted flag against challenge.yml flags; no tolerance
47
  Companions (not included): challenge.yml; README.md
48
  Dependencies: challenge.yml and README.md hold the flag and the official solution path (hidden from the solver); solver tools are free choices (for example a pcap, zip, image or PDF tool), version fixed when the pool is built
campaign_files/2026-10-sc-science-reference-data-2/wfa2-lib-utest/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 4 of 4 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_output_file. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Give the agent wfa.utest.seq and ask for the edit distance (unit cost mismatches, insertions and deletions) of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a edit-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent; give the agent wfa.utest.seq and ask for the optimal gap-affine alignment score under mismatch 4, gap opening 6 and gap extension 2, written as a negative number of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a gap-affine-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent; give the agent wfa.utest.seq and ask for the optimal two-piece gap-affine alignment score under mismatch 4, gap pieces (6, 2) and (24, 1), written as a negative number of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a gap-affine2p-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent; give the agent wfa.utest.seq and ask for the indel distance (minimum number of inserted and deleted bases, mismatches not allowed) of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a indel-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent.
14
 
15
  **Licence (SPDX):** MIT
@@ -21,32 +19,24 @@ Files:
21
  - `tests/wfa.utest.check/test.score.edit.alg`: WFA2-lib unit test: Levenshtein edit distance of 305 synthetic DNA pattern and text pairs (score-only mode).
22
  Format: expected score file, 305 lines (one '<score>\t-' line per pair); input wfa.utest.seq of 610 lines (355811 bytes): 305 synthetic DNA pattern and text pairs ('>' and '<' lines) of 5 to 10000 bases
23
  Origin: test fixture of the upstream test suite (tool_fixture): input written for the case, expected output produced by the tool and committed as the regression reference
24
- Reference result: expected_output_file
25
- Reference source: tests/wfa.utest.check/test.score.edit.alg, the reference file wfa.utest.cmp.sh diffs the align_benchmark output test.score.edit.alg against (wfa.utest.sh line 36 sets --wfa-score-only, line 48 sets -a edit-wfa)
26
  Check: give the agent wfa.utest.seq and ask for the edit distance (unit cost mismatches, insertions and deletions) of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a edit-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent
27
  Companions (not included): tests/wfa.utest.seq; tests/wfa.utest.sh; tests/wfa.utest.cmp.sh; tools/align_benchmark/align_benchmark_params.c; LICENSE
28
  Dependencies: WFA2-lib at 3282f834713cb9767884949143fd0c803acae697; GNU diffutils diff and awk for wfa.utest.cmp.sh
29
  - `tests/wfa.utest.check/test.score.affine.alg`: WFA2-lib unit test: optimal gap-affine alignment score (mismatch 4, gap open 6, extend 2) of 305 synthetic DNA pairs.
30
  Format: expected score file, 305 lines (one '<score>\t-' line per pair); input wfa.utest.seq of 610 lines (355811 bytes): 305 synthetic DNA pattern and text pairs ('>' and '<' lines) of 5 to 10000 bases
31
  Origin: test fixture of the upstream test suite (tool_fixture): input written for the case, expected output produced by the tool and committed as the regression reference
32
- Reference result: expected_output_file
33
- Reference source: tests/wfa.utest.check/test.score.affine.alg, the reference file wfa.utest.cmp.sh diffs the align_benchmark output test.score.affine.alg against (wfa.utest.sh line 36 sets --wfa-score-only, line 49 sets -a gap-affine-wfa)
34
  Check: give the agent wfa.utest.seq and ask for the optimal gap-affine alignment score under mismatch 4, gap opening 6 and gap extension 2, written as a negative number of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a gap-affine-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent
35
  Companions (not included): tests/wfa.utest.seq; tests/wfa.utest.sh; tests/wfa.utest.cmp.sh; tools/align_benchmark/align_benchmark_params.c; LICENSE
36
  Dependencies: WFA2-lib at 3282f834713cb9767884949143fd0c803acae697; GNU diffutils diff and awk for wfa.utest.cmp.sh
37
  - `tests/wfa.utest.check/test.score.affine2p.alg`: WFA2-lib unit test: optimal dual-cost gap-affine alignment score of 305 synthetic DNA pairs.
38
  Format: expected score file, 305 lines (one '<score>\t-' line per pair); input wfa.utest.seq of 610 lines (355811 bytes): 305 synthetic DNA pattern and text pairs ('>' and '<' lines) of 5 to 10000 bases
39
  Origin: test fixture of the upstream test suite (tool_fixture): input written for the case, expected output produced by the tool and committed as the regression reference
40
- Reference result: expected_output_file
41
- Reference source: tests/wfa.utest.check/test.score.affine2p.alg, the reference file wfa.utest.cmp.sh diffs the align_benchmark output test.score.affine2p.alg against (wfa.utest.sh line 36 sets --wfa-score-only, line 50 sets -a gap-affine2p-wfa)
42
  Check: give the agent wfa.utest.seq and ask for the optimal two-piece gap-affine alignment score under mismatch 4, gap pieces (6, 2) and (24, 1), written as a negative number of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a gap-affine2p-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent
43
  Companions (not included): tests/wfa.utest.seq; tests/wfa.utest.sh; tests/wfa.utest.cmp.sh; tools/align_benchmark/align_benchmark_params.c; LICENSE
44
  Dependencies: WFA2-lib at 3282f834713cb9767884949143fd0c803acae697; GNU diffutils diff and awk for wfa.utest.cmp.sh
45
  - `tests/wfa.utest.check/test.score.indel.alg`: WFA2-lib unit test: indel distance (insertions and deletions only) of 305 synthetic DNA pattern and text pairs.
46
  Format: expected score file, 305 lines (one '<score>\t-' line per pair); input wfa.utest.seq of 610 lines (355811 bytes): 305 synthetic DNA pattern and text pairs ('>' and '<' lines) of 5 to 10000 bases
47
  Origin: test fixture of the upstream test suite (tool_fixture): input written for the case, expected output produced by the tool and committed as the regression reference
48
- Reference result: expected_output_file
49
- Reference source: tests/wfa.utest.check/test.score.indel.alg, the reference file wfa.utest.cmp.sh diffs the align_benchmark output test.score.indel.alg against (wfa.utest.sh line 36 sets --wfa-score-only, line 47 sets -a indel-wfa)
50
  Check: give the agent wfa.utest.seq and ask for the indel distance (minimum number of inserted and deleted bases, mismatches not allowed) of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a indel-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent
51
  Companions (not included): tests/wfa.utest.seq; tests/wfa.utest.sh; tests/wfa.utest.cmp.sh; tools/align_benchmark/align_benchmark_params.c; LICENSE
52
  Dependencies: WFA2-lib at 3282f834713cb9767884949143fd0c803acae697; GNU diffutils diff and awk for wfa.utest.cmp.sh
 
8
 
9
  **Self-contained:** Not self-contained: each file is used together with other files from the same upstream. 4 of 4 files name companion files (other upstream files read alongside); companions are not included in this delivery and are listed per file in the notes. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Give the agent wfa.utest.seq and ask for the edit distance (unit cost mismatches, insertions and deletions) of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a edit-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent; give the agent wfa.utest.seq and ask for the optimal gap-affine alignment score under mismatch 4, gap opening 6 and gap extension 2, written as a negative number of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a gap-affine-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent; give the agent wfa.utest.seq and ask for the optimal two-piece gap-affine alignment score under mismatch 4, gap pieces (6, 2) and (24, 1), written as a negative number of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a gap-affine2p-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent; give the agent wfa.utest.seq and ask for the indel distance (minimum number of inserted and deleted bases, mismatches not allowed) of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a indel-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent.
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `tests/wfa.utest.check/test.score.edit.alg`: WFA2-lib unit test: Levenshtein edit distance of 305 synthetic DNA pattern and text pairs (score-only mode).
20
  Format: expected score file, 305 lines (one '<score>\t-' line per pair); input wfa.utest.seq of 610 lines (355811 bytes): 305 synthetic DNA pattern and text pairs ('>' and '<' lines) of 5 to 10000 bases
21
  Origin: test fixture of the upstream test suite (tool_fixture): input written for the case, expected output produced by the tool and committed as the regression reference
 
 
22
  Check: give the agent wfa.utest.seq and ask for the edit distance (unit cost mismatches, insertions and deletions) of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a edit-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent
23
  Companions (not included): tests/wfa.utest.seq; tests/wfa.utest.sh; tests/wfa.utest.cmp.sh; tools/align_benchmark/align_benchmark_params.c; LICENSE
24
  Dependencies: WFA2-lib at 3282f834713cb9767884949143fd0c803acae697; GNU diffutils diff and awk for wfa.utest.cmp.sh
25
  - `tests/wfa.utest.check/test.score.affine.alg`: WFA2-lib unit test: optimal gap-affine alignment score (mismatch 4, gap open 6, extend 2) of 305 synthetic DNA pairs.
26
  Format: expected score file, 305 lines (one '<score>\t-' line per pair); input wfa.utest.seq of 610 lines (355811 bytes): 305 synthetic DNA pattern and text pairs ('>' and '<' lines) of 5 to 10000 bases
27
  Origin: test fixture of the upstream test suite (tool_fixture): input written for the case, expected output produced by the tool and committed as the regression reference
 
 
28
  Check: give the agent wfa.utest.seq and ask for the optimal gap-affine alignment score under mismatch 4, gap opening 6 and gap extension 2, written as a negative number of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a gap-affine-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent
29
  Companions (not included): tests/wfa.utest.seq; tests/wfa.utest.sh; tests/wfa.utest.cmp.sh; tools/align_benchmark/align_benchmark_params.c; LICENSE
30
  Dependencies: WFA2-lib at 3282f834713cb9767884949143fd0c803acae697; GNU diffutils diff and awk for wfa.utest.cmp.sh
31
  - `tests/wfa.utest.check/test.score.affine2p.alg`: WFA2-lib unit test: optimal dual-cost gap-affine alignment score of 305 synthetic DNA pairs.
32
  Format: expected score file, 305 lines (one '<score>\t-' line per pair); input wfa.utest.seq of 610 lines (355811 bytes): 305 synthetic DNA pattern and text pairs ('>' and '<' lines) of 5 to 10000 bases
33
  Origin: test fixture of the upstream test suite (tool_fixture): input written for the case, expected output produced by the tool and committed as the regression reference
 
 
34
  Check: give the agent wfa.utest.seq and ask for the optimal two-piece gap-affine alignment score under mismatch 4, gap pieces (6, 2) and (24, 1), written as a negative number of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a gap-affine2p-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent
35
  Companions (not included): tests/wfa.utest.seq; tests/wfa.utest.sh; tests/wfa.utest.cmp.sh; tools/align_benchmark/align_benchmark_params.c; LICENSE
36
  Dependencies: WFA2-lib at 3282f834713cb9767884949143fd0c803acae697; GNU diffutils diff and awk for wfa.utest.cmp.sh
37
  - `tests/wfa.utest.check/test.score.indel.alg`: WFA2-lib unit test: indel distance (insertions and deletions only) of 305 synthetic DNA pattern and text pairs.
38
  Format: expected score file, 305 lines (one '<score>\t-' line per pair); input wfa.utest.seq of 610 lines (355811 bytes): 305 synthetic DNA pattern and text pairs ('>' and '<' lines) of 5 to 10000 bases
39
  Origin: test fixture of the upstream test suite (tool_fixture): input written for the case, expected output produced by the tool and committed as the regression reference
 
 
40
  Check: give the agent wfa.utest.seq and ask for the indel distance (minimum number of inserted and deleted bases, mismatches not allowed) of every pattern and text pair, one line per pair in input order written as the score, a tab and '-', as the test does with align_benchmark -i wfa.utest.seq -a indel-wfa --wfa-score-only; pass when the score file is identical to the expected file; the expected file is the answer and is hidden from the agent
41
  Companions (not included): tests/wfa.utest.seq; tests/wfa.utest.sh; tests/wfa.utest.cmp.sh; tools/align_benchmark/align_benchmark_params.c; LICENSE
42
  Dependencies: WFA2-lib at 3282f834713cb9767884949143fd0c803acae697; GNU diffutils diff and awk for wfa.utest.cmp.sh
campaign_files/2026-10-sw-tests-and-exercises-2/wren-test-suite/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** Whether the file needs other upstream files is not stated. Dependencies named upstream are listed per file in the notes.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_output. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Run the file with wren_test and compare stdout line by line with the 25 expect lines in order; exact match, exit code 0.
14
 
15
  **Licence (SPDX):** MIT
@@ -21,7 +19,5 @@ Files:
21
  - `test/core/string/index_of_start.wren`: Wren language test suite, test/core/string/index_of_start.wren: String.indexOf(_,_) with 25 expected results in // expect: comments.
22
  Format: Wren source file; expected stdout lines as // expect: <text> comments
23
  Origin: tool_fixture: an official test case of the project's own conformance or regression suite, with its expected result
24
- Reference result: expected_output
25
- Reference source: the // expect: comments of the unit in order; util/test.py line 40 of the same commit reads them with EXPECT_PATTERN = re.compile(r'// expect: ?(.*)')
26
  Check: run the file with wren_test and compare stdout line by line with the 25 expect lines in order; exact match, exit code 0
27
  Dependencies: a Wren VM under test (wren_test from the same repository; latest release before the commit date 2025-11-19 is 0.4.0 of 2021-04-09, read from api.github.com/repos/wren-lang/wren/releases)
 
8
 
9
  **Self-contained:** Whether the file needs other upstream files is not stated. Dependencies named upstream are listed per file in the notes.
10
 
 
 
11
  **Tools to open and check:** Check: Run the file with wren_test and compare stdout line by line with the 25 expect lines in order; exact match, exit code 0.
12
 
13
  **Licence (SPDX):** MIT
 
19
  - `test/core/string/index_of_start.wren`: Wren language test suite, test/core/string/index_of_start.wren: String.indexOf(_,_) with 25 expected results in // expect: comments.
20
  Format: Wren source file; expected stdout lines as // expect: <text> comments
21
  Origin: tool_fixture: an official test case of the project's own conformance or regression suite, with its expected result
 
 
22
  Check: run the file with wren_test and compare stdout line by line with the 25 expect lines in order; exact match, exit code 0
23
  Dependencies: a Wren VM under test (wren_test from the same repository; latest release before the commit date 2025-11-19 is 0.4.0 of 2021-04-09, read from api.github.com/repos/wren-lang/wren/releases)
campaign_files/2026-10-sw-tests-and-exercises/wpt-url-testdata/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is marked self-contained upstream.
10
 
11
- **Answers:** No answer files are included. Reference result: expected_url_components. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Parse input against base with the URL parser under test; compare href and each component, or require failure. One array element is one case.
14
 
15
  **Licence (SPDX):** BSD-3-Clause
@@ -21,6 +19,4 @@ Files:
21
  - `url/resources/urltestdata.json`: Web-platform-tests URL parsing data urltestdata.json.
22
  Format: JSON array of {input, base, href, protocol, host, ...} or {failure: true}
23
  Origin: official conformance test suite shipped by the project or standards body (was official_conformance_suite)
24
- Reference result: expected_url_components
25
- Reference source: each object gives the expected serialized href and components, or failure: true when parsing must fail
26
  Check: Parse input against base with the URL parser under test; compare href and each component, or require failure. One array element is one case.
 
8
 
9
  **Self-contained:** The file is marked self-contained upstream.
10
 
 
 
11
  **Tools to open and check:** Check: Parse input against base with the URL parser under test; compare href and each component, or require failure. One array element is one case.
12
 
13
  **Licence (SPDX):** BSD-3-Clause
 
19
  - `url/resources/urltestdata.json`: Web-platform-tests URL parsing data urltestdata.json.
20
  Format: JSON array of {input, base, href, protocol, host, ...} or {failure: true}
21
  Origin: official conformance test suite shipped by the project or standards body (was official_conformance_suite)
 
 
22
  Check: Parse input against base with the URL parser under test; compare href and each component, or require failure. One array element is one case.
campaign_files/2026-10-sw-tests-and-exercises/wycheproof/README.md CHANGED
@@ -8,8 +8,6 @@
8
 
9
  **Self-contained:** The file is marked self-contained upstream.
10
 
11
- **Answers:** No answer files are included. Reference result: valid_invalid_verdict. Where the reference result comes from and how a solution is checked is written per file in the notes.
12
-
13
  **Tools to open and check:** Check: Run encrypt/decrypt with the AES-GCM implementation under test; valid cases must reproduce ct and tag, invalid cases must be rejected; acceptable cases are reported only.
14
 
15
  **Licence (SPDX):** Apache-2.0
@@ -21,6 +19,4 @@ Files:
21
  - `testvectors_v1/aes_gcm_test.json`: Wycheproof AES-GCM test vectors (testvectors_v1/aes_gcm_test.json).
22
  Format: JSON {testGroups:[{tests:[{tcId, key, iv, aad, msg, ct, tag, result, flags}]}]}
23
  Origin: official conformance test suite shipped by the project or standards body (was official_conformance_suite)
24
- Reference result: valid_invalid_verdict
25
- Reference source: each test has result valid, invalid or acceptable plus expected ct and tag
26
  Check: Run encrypt/decrypt with the AES-GCM implementation under test; valid cases must reproduce ct and tag, invalid cases must be rejected; acceptable cases are reported only.
 
8
 
9
  **Self-contained:** The file is marked self-contained upstream.
10
 
 
 
11
  **Tools to open and check:** Check: Run encrypt/decrypt with the AES-GCM implementation under test; valid cases must reproduce ct and tag, invalid cases must be rejected; acceptable cases are reported only.
12
 
13
  **Licence (SPDX):** Apache-2.0
 
19
  - `testvectors_v1/aes_gcm_test.json`: Wycheproof AES-GCM test vectors (testvectors_v1/aes_gcm_test.json).
20
  Format: JSON {testGroups:[{tests:[{tcId, key, iv, aad, msg, ct, tag, result, flags}]}]}
21
  Origin: official conformance test suite shipped by the project or standards body (was official_conformance_suite)
 
 
22
  Check: Run encrypt/decrypt with the AES-GCM implementation under test; valid cases must reproduce ct and tag, invalid cases must be rejected; acceptable cases are reported only.
cards/0xdea-semgrep-rules.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = '0xdea-semgrep-rules'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = '0xdea-semgrep-rules'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = '0xdea-semgrep-rules'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = '0xdea-semgrep-rules'
cards/1brc-test-samples.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = '1brc-test-samples'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = '1brc-test-samples'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = '1brc-test-samples'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = '1brc-test-samples'
cards/2dpacklib.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = '2dpacklib'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = '2dpacklib'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = '2dpacklib'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = '2dpacklib'
cards/abacus-integrate-tests.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'abacus-integrate-tests'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'abacus-integrate-tests'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'abacus-integrate-tests'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'abacus-integrate-tests'
cards/abhishekumarr00-supply-chain-mastery.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'abhishekumarr00-supply-chain-mastery'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'abhishekumarr00-supply-chain-mastery'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'abhishekumarr00-supply-chain-mastery'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'abhishekumarr00-supply-chain-mastery'
cards/academy-bus-nj-gtfs.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'academy-bus-nj-gtfs'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'academy-bus-nj-gtfs'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'academy-bus-nj-gtfs'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'academy-bus-nj-gtfs'
cards/accordproject-cicero-template-library.md CHANGED
@@ -52,10 +52,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
52
 
53
  ## Query this source at a fixed release
54
 
55
- Pinned to the release tag `release-2026-11-12.9`.
56
 
57
  ```sql
58
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'accordproject-cicero-template-library'));
59
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
60
  FROM read_parquet(getvariable('files'))
61
  WHERE source = 'accordproject-cicero-template-library'
 
52
 
53
  ## Query this source at a fixed release
54
 
55
+ Pinned to the release tag `release-2026-11-12.10`.
56
 
57
  ```sql
58
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'accordproject-cicero-template-library'));
59
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
60
  FROM read_parquet(getvariable('files'))
61
  WHERE source = 'accordproject-cicero-template-library'
cards/activity-schema-aggregate-in-between-1.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'activity-schema-aggregate-in-between-1'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'activity-schema-aggregate-in-between-1'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'activity-schema-aggregate-in-between-1'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'activity-schema-aggregate-in-between-1'
cards/activity-schema-first-in-between-2.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'activity-schema-first-in-between-2'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'activity-schema-first-in-between-2'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'activity-schema-first-in-between-2'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'activity-schema-first-in-between-2'
cards/acwf-verification.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'acwf-verification'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'acwf-verification'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'acwf-verification'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'acwf-verification'
cards/adflib-adf-deleted.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'adflib-adf-deleted'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'adflib-adf-deleted'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'adflib-adf-deleted'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'adflib-adf-deleted'
cards/adflib-adf-malformed.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'adflib-adf-malformed'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'adflib-adf-malformed'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'adflib-adf-malformed'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'adflib-adf-malformed'
cards/adventure-works-oltp-sample.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'adventure-works-oltp-sample'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'adventure-works-oltp-sample'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'adventure-works-oltp-sample'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'adventure-works-oltp-sample'
cards/agat-tests.md CHANGED
@@ -52,10 +52,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
52
 
53
  ## Query this source at a fixed release
54
 
55
- Pinned to the release tag `release-2026-11-12.9`.
56
 
57
  ```sql
58
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'agat-tests'));
59
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
60
  FROM read_parquet(getvariable('files'))
61
  WHERE source = 'agat-tests'
 
52
 
53
  ## Query this source at a fixed release
54
 
55
+ Pinned to the release tag `release-2026-11-12.10`.
56
 
57
  ```sql
58
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'agat-tests'));
59
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
60
  FROM read_parquet(getvariable('files'))
61
  WHERE source = 'agat-tests'
cards/agl-gerrit-changes.md CHANGED
@@ -50,10 +50,10 @@ Size quantiles: unknown (no size distribution recorded for this source).
50
 
51
  ## Query this source at a fixed release
52
 
53
- Pinned to the release tag `release-2026-11-12.9`.
54
 
55
  ```sql
56
- SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.9/meta/catalog.parquet' WHERE source = 'agl-gerrit-changes'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'agl-gerrit-changes'
 
50
 
51
  ## Query this source at a fixed release
52
 
53
+ Pinned to the release tag `release-2026-11-12.10`.
54
 
55
  ```sql
56
+ SET VARIABLE files = (SELECT list('hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/' || f) FROM (SELECT unnest(data_files) AS f FROM 'hf://datasets/Battam/training-data-foundation@release-2026-11-12.10/meta/catalog.parquet' WHERE source = 'agl-gerrit-changes'));
57
  SELECT uri, ext, size_bytes, lower(hex(sha256)) AS sha256
58
  FROM read_parquet(getvariable('files'))
59
  WHERE source = 'agl-gerrit-changes'