# tasksource licence audit — exact rows used in Ines-1 Factual dossier for licence review. **It does not conclude whether the model weights are a derivative work and does not choose a licence.** **Scope:** the 24,959 tasksource questions that actually entered Decision-Mix Stage 2 (`train_tasksource.jsonl` of the mix), joined row by row with `tasksource/tasksource-jev-typed-decisions` @ `8173a06c` (24,959 of 24,959 matched, 0 missing), and with that revision's `sources.yaml` (tasksource commit `b38efb66`), which records for each source the Hub dataset and revision it was loaded from, the original dataset behind a tasksource copy, the dataset-card licences and the licences recorded by the Data Provenance Initiative (DPI). Full table: [tasksource_license_audit.csv](tasksource_license_audit.csv) (one row per unique original source). Selection rule used when building the mix: `license_use == "commercial"` only, at most 80 questions per source, no source derived from typed-decisions or jev-decisions. Per the tasksource card, `license_use` is `commercial` when one of the recorded licences allows commercial use, **share-alike and copyleft included**, and the licence metadata is "a best-effort aid, not legal advice. Licenses on cards can be wrong or incomplete, and a source's terms may differ from its card's." **Category** = most restrictive class among the source's recorded licences, in the order copyleft (GPL, AGPL, LGPL, MPL) > share-alike (CC-BY-SA) > database licence (ODbL, ODC-By) > unclear/other (`other`, `Custom`, `OANC`, bare `cc`) > attribution (CC-BY) > permissive (MIT, Apache-2.0, BSD, AFL-3.0, CC0). Other classes present are in *notes*. ## Summary - Unique sources used: **366** - Questions: **24,959** | category | sources | questions | |---|---:|---:| | copyleft | 10 | 690 | | share-alike | 73 | 5,107 | | database licence | 5 | 346 | | unclear/other | 10 | 685 | | attribution | 85 | 5,481 | | permissive | 183 | 12,650 | Questions per recorded licence string token (a source with several licences counts under each): | licence | questions | |---|---:| | mit | 5,178 | | cc-by-4.0 | 5,119 | | apache-2.0 | 4,053 | | CC BY-SA 4.0 (DPI) | 2,950 | | cc-by-sa-4.0 | 2,783 | | bsd | 1,394 | | CC BY 4.0 (DPI) | 1,183 | | afl-3.0 | 1,175 | | OANC (DPI) | 1,112 | | MIT License (DPI) | 1,044 | | cc0-1.0 | 703 | | Apache License 2.0 (DPI) | 623 | | Custom (DPI) | 556 | | cc-by-sa-3.0 | 349 | | gpl-3.0 | 281 | | odc-by | 278 | | CC0 1.0 (DPI) | 278 | | gpl | 274 | | cc | 146 | | BSD 2-Clause License (DPI) | 143 | | cc-by-3.0 | 139 | | mpl-2.0 | 69 | | CC BY-SA 3.0 (DPI) | 69 | | odbl | 68 | | other | 67 | | agpl-3.0 | 66 | ### GPL sources - `dynahate` (67 questions; gpl) - `fever-evidence-related` (72 questions; cc-by-sa-3.0, gpl-3.0, CC BY-SA 4.0 (DPI)) - `implicatures` (67 questions; gpl) - `multilingual/universal-joy` (70 questions; gpl) - `scientific-exaggeration-detection` (65 questions; gpl-3.0) - `tomi-nli` (74 questions; gpl-3.0) - `TroFi` (70 questions; gpl, Custom (DPI)) - `tweets_hate_speech_detection` (70 questions; gpl-3.0) ### AGPL sources - `ethos/multilabel` (66 questions; agpl-3.0) ### LGPL sources - none ### CC-BY-SA sources - `ai2_arc/ARC-Challenge/challenge` (68 questions; cc-by-sa-4.0) - `ai2_arc/ARC-Easy/challenge` (71 questions; cc-by-sa-4.0) - `breaking_nli` (64 questions; CC BY-SA 4.0 (DPI)) - `CREAK` (70 questions; CC BY-SA 4.0 (DPI)) - `dbpedia_14/dbpedia_14` (71 questions; cc-by-sa-3.0) - `fever-evidence-related` (72 questions; cc-by-sa-3.0, gpl-3.0, CC BY-SA 4.0 (DPI)) - `folio` (69 questions; cc, CC BY-SA 4.0 (DPI)) - `gen_debiased_nli/mnli_par_z` (68 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/mnli_seq_z` (73 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/mnli_z_aug` (67 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/snli_par_z` (73 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/snli_seq_z` (74 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/snli_z_aug` (70 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `google_wellformed_query` (66 questions; cc-by-sa-4.0) - `hate_speech18` (69 questions; cc-by-sa-3.0, CC BY-SA 3.0 (DPI)) - `help-nli` (77 questions; cc, CC BY-SA 4.0 (DPI)) - `MedQA-USMLE-4-options-hf` (72 questions; cc-by-sa-4.0) - `MSciNLI` (68 questions; cc-by-sa-4.0) - `multilingual/americas_nli/all_languages` (70 questions; cc-by-sa-4.0) - `multilingual/exams/multilingual` (71 questions; cc-by-sa-4.0) - `multilingual/klue/nli` (71 questions; cc-by-sa-4.0) - `multilingual/klue/sts` (70 questions; cc-by-sa-4.0) - `multilingual/klue/ynat` (74 questions; cc-by-sa-4.0) - `multilingual/miam` (63 questions; cc-by-sa-4.0) - `multilingual/mtop` (67 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/ace` (71 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/ban` (65 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/bbc` (67 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/bjn` (74 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/bug` (71 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/eng` (73 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/ind` (70 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/jav` (70 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/mad` (69 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/min` (71 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/nij` (66 questions; cc-by-sa-4.0) - `multilingual/NusaX-senti/sun` (70 questions; cc-by-sa-4.0) - `multilingual/xstory_cloze/ar` (73 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/en` (64 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/es` (71 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/eu` (66 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/hi` (72 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/id` (65 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/my` (76 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/ru` (69 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/sw` (73 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/te` (60 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `multilingual/xstory_cloze/zh` (70 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `nan-nli` (74 questions; cc-by-sa-4.0) - `onestop_qa` (72 questions; cc-by-sa-4.0, CC BY-SA 4.0 (DPI)) - `probability_words_nli/reasoning_1hop` (76 questions; Apache License 2.0 (DPI), CC BY-SA 4.0 (DPI)) - `probability_words_nli/reasoning_2hop` (69 questions; Apache License 2.0 (DPI), CC BY-SA 4.0 (DPI)) - `probability_words_nli/usnli` (64 questions; Apache License 2.0 (DPI), CC BY-SA 4.0 (DPI)) - `recast/recast_factuality` (74 questions; CC BY-SA 4.0 (DPI)) - `recast/recast_megaveridicality` (74 questions; CC BY-SA 4.0 (DPI)) - `recast/recast_ner` (72 questions; CC BY-SA 4.0 (DPI)) - `recast/recast_puns` (73 questions; CC BY-SA 4.0 (DPI)) - `recast/recast_sentiment` (73 questions; CC BY-SA 4.0 (DPI)) - `recast/recast_verbcorner` (71 questions; CC BY-SA 4.0 (DPI)) - `recast/recast_verbnet` (70 questions; CC BY-SA 4.0 (DPI)) - `robust_nli/IS_CS` (74 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/LI_LI` (74 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/PI_CD` (67 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/PI_SP` (65 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/ST_LM` (68 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/ST_NE` (70 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/ST_SE` (69 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/ST_WO` (68 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `scinli` (71 questions; apache-2.0, CC BY-SA 4.0 (DPI)) - `silicone/maptask` (67 questions; cc-by-sa-4.0, CC BY 4.0 (DPI)) - `snli` (69 questions; cc-by-sa-4.0) - `vitaminc` (70 questions; cc-by-sa-3.0) - `wice` (74 questions; cc-by-sa-4.0) - `wiki_hop/original` (67 questions; cc-by-sa-3.0) ### ODbL sources - `multilingual/wili_2018` (68 questions; odbl) ### ODC-By sources - `codah/codah` (72 questions; odc-by) - `wildguardmix-cleaned/prompt_harm` (72 questions; odc-by) - `wildguardmix-cleaned/response_harm` (65 questions; odc-by) - `wildguardmix-cleaned/response_refusal` (69 questions; odc-by) ### MPL sources - `emo/emo2019` (69 questions; mpl-2.0) ### `other` / undetermined licences (`other`, `Custom (DPI)`, `OANC (DPI)`, bare `cc`) - `head_qa/en` (67 questions; mit, other) - `blog_authorship_corpus/gender` (70 questions; apache-2.0, Custom (DPI)) - `blog_authorship_corpus/job` (70 questions; apache-2.0, Custom (DPI)) - `conceptrules_v2` (67 questions; mit, Custom (DPI)) - `lex_glue/scotus` (74 questions; cc-by-4.0, Custom (DPI)) - `open_question_type` (69 questions; cc-by-4.0, CC BY 4.0 (DPI), Custom (DPI)) - `rumoureval_2019/RumourEval2019` (63 questions; cc-by-4.0, CC BY 4.0 (DPI), Custom (DPI)) - `scruples/verdict_votes` (73 questions; apache-2.0, Custom (DPI)) - `TroFi` (70 questions; gpl, Custom (DPI)) - `gen_debiased_nli/mnli_par_z` (68 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/mnli_seq_z` (73 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/mnli_z_aug` (67 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/snli_par_z` (73 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/snli_seq_z` (74 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `gen_debiased_nli/snli_z_aug` (70 questions; CC BY-SA 4.0 (DPI), OANC (DPI)) - `robust_nli/IS_CS` (74 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/LI_LI` (74 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/PI_CD` (67 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/PI_SP` (65 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/ST_LM` (68 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/ST_NE` (70 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/ST_SE` (69 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli/ST_WO` (68 questions; CC BY-SA 4.0 (DPI), MIT License (DPI), OANC (DPI)) - `robust_nli_is_sd` (64 questions; OANC (DPI)) - `robust_nli_li_ts` (68 questions; OANC (DPI)) - `folio` (69 questions; cc, CC BY-SA 4.0 (DPI)) - `help-nli` (77 questions; cc, CC BY-SA 4.0 (DPI)) --- Generated from the mix files and the pinned tasksource metadata; no rows or examples are reproduced here.