--- license: other license_name: mixed-source-licences license_link: https://huggingface.co/datasets/a1i6ek/den-datasets/blob/main/LICENSES.md pretty_name: den training and evaluation data tags: [den, system-one, multiple-choice, calibration] --- # den data The training, calibration, development and test data of den (github.com/ramput-labs/den): Kev 1.0's frozen suites, normalized public sources and their cleaned copies. **Each file keeps the licence of the sources it holds**; there is no single licence for the whole. `LICENSES.md` lists every file's sources and class, `LICENSES/` holds each source's terms, the evidence they were read from, the canonical licence texts and the licence files shipped with the data. Some sources are **non-commercial** (CC-BY-NC or research-only): files holding them may be used for non-commercial research only. Share-alike sources (CC-BY-SA) require derived data to keep their licence. Cite the sources below when you use them. ## Sources | source | licence | class | terms | |---|---|---|---| | AART | CC-BY-4.0 | open | [terms](https://huggingface.co/datasets/walledai/AART) · [details](LICENSES/aart.md) | | Aegis AI Content Safety 2.0 | CC-BY-4.0 | open | [terms](https://huggingface.co/datasets/nvidia/Aegis-AI-Content-Safety-Dataset-2.0) · [details](LICENSES/aegis.md) | | AG News | non-commercial academic use | non-commercial | [terms](http://www.di.unipi.it/~gulli/AG_corpus_of_news_articles.html) · [details](LICENSES/ag-news.md) | | AI2 ARC | CC-BY-SA-4.0 | share-alike | [terms](https://huggingface.co/datasets/allenai/ai2_arc) · [details](LICENSES/arc.md) | | API-Bank | MIT | open | [terms](https://huggingface.co/datasets/liminghao1630/API-Bank) · [details](LICENSES/apibank.md) | | BANKING77 | CC-BY-4.0 | open | [terms](https://huggingface.co/datasets/legacy-datasets/banking77) · [details](LICENSES/banking77.md) | | Berkeley Function-Calling Leaderboard | Apache-2.0 | open | [terms](https://huggingface.co/datasets/gorilla-llm/Berkeley-Function-Calling-Leaderboard) · [details](LICENSES/bfcl.md) | | BoolQ | CC-BY-SA-3.0 | share-alike | [terms](https://huggingface.co/datasets/google/boolq) · [details](LICENSES/boolq.md) | | BRIGHT | CC-BY-4.0 | open | [terms](https://huggingface.co/datasets/xlangai/BRIGHT) · [details](LICENSES/bright.md) | | CFPB Consumer Complaint Database | US government work (public domain) | open | [terms](https://www.consumerfinance.gov/data-research/consumer-complaints/) · [details](LICENSES/cfpb.md) | | ChessBench action values | CC-BY-4.0, with lichess positions under CC0 | open | [terms](https://github.com/google-deepmind/searchless_chess) · [details](LICENSES/chessbench.md) | | CLINC150 (OOS) | CC-BY-3.0 | open | [terms](https://huggingface.co/datasets/clinc/clinc_oos) · [details](LICENSES/clinc150.md) | | CodeReviewer (diff quality) | CC-BY-4.0; code under each project's permissive licence | open | [terms](https://zenodo.org/records/6900648) · [details](LICENSES/codereviewer.md) | | Collaborative Filtering of Color Aesthetics | none stated for the data | unspecified | [terms](https://www.dgp.toronto.edu/~donovan/cfcolor/) · [details](LICENSES/cfcolor.md) | | CommitPackFT | MIT (dataset); rows kept only under permissive repository licences | open | [terms](https://huggingface.co/datasets/bigcode/commitpackft) · [details](LICENSES/commitpackft.md) | | CommonsenseQA | MIT | open | [terms](https://huggingface.co/datasets/tau/commonsense_qa) · [details](LICENSES/commonsenseqa.md) | | ContractNLI | CC-BY-4.0 | open | [terms](https://stanfordnlp.github.io/contract-nli/) · [details](LICENSES/contractnli.md) | | dair-ai/emotion | research and educational use only | non-commercial | [terms](https://huggingface.co/datasets/dair-ai/emotion) · [details](LICENSES/emotion.md) | | DBpedia-14 | CC-BY-SA-3.0 | share-alike | [terms](https://huggingface.co/datasets/fancyzhx/dbpedia_14) · [details](LICENSES/dbpedia14.md) | | FlakeFlagger | CC-BY-4.0 (data); BSD-3-Clause (repository); test code under each project's licence | open | [terms](https://zenodo.org/records/4450723) · [details](LICENSES/flakeflagger.md) | | Glaive function calling v2 | Apache-2.0 | open | [terms](https://huggingface.co/datasets/glaiveai/glaive-function-calling-v2) · [details](LICENSES/glaive.md) | | HellaSwag (ActivityNet items) | MIT | open | [terms](https://huggingface.co/datasets/Rowan/hellaswag) · [details](LICENSES/hellaswag.md) | | HellaSwag validation file (raw, with wikiHow items) | MIT, subject to wikiHow's DMCA notice | restricted | [terms](https://github.com/github/dmca/blob/master/2026/09/2026-09-14-wikihow.md) · [details](LICENSES/hellaswag-raw.md) | | Humicroedit (SemEval-2020 Task 7) | none stated (cite the task paper) | unspecified | [terms](https://cs.rochester.edu/u/nhossain/humicroedit.html) · [details](LICENSES/humicroedit.md) | | Kev 1.0 generated data | Apache-2.0 | open | [terms](https://github.com/jaredpalmer/kev/blob/fe64b1274ea7f80d4095866df90666abb03e9cf6/LICENSE) · [details](LICENSES/kev.md) | | Large Movie Review Dataset (IMDB) | none stated | unspecified | [terms](https://ai.stanford.edu/~amaas/data/sentiment/) · [details](LICENSES/imdb.md) | | MASSIVE (en-US) | CC-BY-4.0 | open | [terms](https://github.com/alexa/massive/blob/main/NOTICE.md) · [details](LICENSES/massive.md) | | MMLU | MIT | open | [terms](https://huggingface.co/datasets/cais/mmlu) · [details](LICENSES/mmlu.md) | | MMLU-Pro | MIT | open | [terms](https://huggingface.co/datasets/TIGER-Lab/MMLU-Pro) · [details](LICENSES/mmlu-pro.md) | | Multilingual Amazon Reviews Corpus (English) | Amazon Reviews Corpus licence (academic research only, no republication) | restricted | [terms](https://huggingface.co/datasets/defunct-datasets/amazon_reviews_multi) · [details](LICENSES/amazon.md) | | MultiNLI | OANC licence; fiction under CC-BY-SA-3.0 / CC-BY-3.0 | share-alike | [terms](https://huggingface.co/datasets/nyu-mll/multi_nli) · [details](LICENSES/mnli.md) | | MuSR | CC-BY-4.0 | open | [terms](https://huggingface.co/datasets/TAUR-Lab/MuSR) · [details](LICENSES/musr.md) | | OpenBookQA | Apache-2.0 | open | [terms](https://github.com/allenai/OpenBookQA) · [details](LICENSES/openbookqa.md) | | PAWS (Wiki) | free use, acknowledgement appreciated | open | [terms](https://huggingface.co/datasets/google-research-datasets/paws) · [details](LICENSES/paws.md) | | Prompt injections (deepset, Lakera Gandalf) | Apache-2.0 (deepset); MIT (Lakera) | open | [terms](https://huggingface.co/datasets/deepset/prompt-injections) · [details](LICENSES/prompt-injection.md) | | QASC | CC-BY-4.0 | open | [terms](https://huggingface.co/datasets/allenai/qasc) · [details](LICENSES/qasc.md) | | QNLI (GLUE, from SQuAD) | CC-BY-SA-4.0 (SQuAD) | share-alike | [terms](https://rajpurkar.github.io/SQuAD-explorer/) · [details](LICENSES/qnli.md) | | RouterBench (MMLU, Winogrande, MBPP items) | MIT | open | [terms](https://github.com/withmartian/routerbench) · [details](LICENSES/routerbench.md) | | RouterBench 0-shot file (raw) | MIT, embedding third-party prompts incl. HellaSwag wikiHow items | restricted | [terms](https://github.com/github/dmca/blob/master/2026/09/2026-09-14-wikihow.md) · [details](LICENSES/routerbench-raw.md) | | SATA-Bench | CC-BY-NC-4.0 | non-commercial | [terms](https://huggingface.co/datasets/sata-bench/sata-bench) · [details](LICENSES/sata-bench.md) | | Schema-Guided Dialogue (+ SGD-X) | CC-BY-SA-4.0 | share-alike | [terms](https://github.com/google-research-datasets/dstc8-schema-guided-dialogue) · [details](LICENSES/sgd.md) | | SciQ | CC-BY-NC-3.0 | non-commercial | [terms](https://huggingface.co/datasets/allenai/sciq) · [details](LICENSES/sciq.md) | | SemIf | MIT | open | [terms](https://github.com/TheoLeeCJ/SemIf) · [details](LICENSES/semif.md) | | Stanford Sentiment Treebank (SST-5) | none stated | unspecified | [terms](https://nlp.stanford.edu/sentiment/) · [details](LICENSES/sst5.md) | | ToolRet | Apache-2.0 (release); queries and tools re-published from their source benchmarks | open | [terms](https://github.com/mangopy/tool-retrieval-benchmark) · [details](LICENSES/toolret.md) | | TREC question classification | none stated | unspecified | [terms](https://cogcomp.seas.upenn.edu/Data/QA/QC/) · [details](LICENSES/trec.md) | | TweetEval offensive (OLID) | none stated; Twitter/X Terms of Service apply | unspecified | [terms](https://huggingface.co/datasets/cardiffnlp/tweet_eval) · [details](LICENSES/tweeteval-offensive.md) | | When2Call | CC-BY-4.0 | open | [terms](https://huggingface.co/datasets/nvidia/When2Call) · [details](LICENSES/when2call.md) | | Yelp Review Full | Yelp Dataset Terms of Use | restricted | [terms](https://s3-media3.fl.yelpcdn.com/assets/srv0/engineering_pages/bea5c1e92bf3/assets/vendor/yelp-dataset-agreement.pdf) · [details](LICENSES/yelp.md) | ## Not in this copy Files whose sources may not be redistributed (or have no stated licence, in a public copy) were left out; rebuild them from the pinned originals with `make data data-raw-train data-raw-new data-raw-eval breadth normalize clean-data`: - `calibration/core.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English), Yelp Review Full - `calibration/heldout.jsonl`: unspecified: TweetEval offensive (OLID) - `clean/calibration/core.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English), Yelp Review Full - `clean/calibration/heldout.jsonl`: unspecified: TweetEval offensive (OLID) - `clean/dev/breadth.jsonl`: unspecified: Collaborative Filtering of Color Aesthetics, Humicroedit (SemEval-2020 Task 7) - `clean/dev/core.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English), Yelp Review Full - `clean/dev/probes.jsonl`: unspecified: TweetEval offensive (OLID) - `clean/dev/sources/sentiment/amazon-reviews.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `clean/dev/sources/sentiment/imdb.jsonl`: unspecified: Large Movie Review Dataset (IMDB) - `clean/dev/sources/sentiment/sst5.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `clean/dev/sources/sentiment/yelp-full.jsonl`: restricted: Yelp Review Full - `clean/dev/sources/topic/trec.jsonl`: unspecified: TREC question classification - `clean/dev/sources/transfer/tweeteval-offensive.jsonl`: unspecified: TweetEval offensive (OLID) - `clean/dev/transfer.jsonl`: unspecified: TweetEval offensive (OLID) - `clean/test/breadth.jsonl`: unspecified: Collaborative Filtering of Color Aesthetics, Humicroedit (SemEval-2020 Task 7) - `clean/test/core.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English), Yelp Review Full - `clean/test/probes.jsonl`: unspecified: TweetEval offensive (OLID) - `clean/test/sources/sentiment/amazon-reviews.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `clean/test/sources/sentiment/imdb.jsonl`: unspecified: Large Movie Review Dataset (IMDB) - `clean/test/sources/sentiment/sst5.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `clean/test/sources/sentiment/yelp-full.jsonl`: restricted: Yelp Review Full - `clean/test/sources/topic/trec.jsonl`: unspecified: TREC question classification - `clean/test/sources/transfer/tweeteval-offensive.jsonl`: unspecified: TweetEval offensive (OLID) - `clean/test/transfer.jsonl`: unspecified: TweetEval offensive (OLID) - `clean/train/core.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English), Yelp Review Full - `clean/train/sources/sentiment/amazon-reviews.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `clean/train/sources/sentiment/imdb.jsonl`: unspecified: Large Movie Review Dataset (IMDB) - `clean/train/sources/sentiment/sst5.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `clean/train/sources/sentiment/yelp-full.jsonl`: restricted: Yelp Review Full - `clean/train/sources/topic/trec.jsonl`: unspecified: TREC question classification - `dev/breadth.jsonl`: unspecified: Collaborative Filtering of Color Aesthetics, Humicroedit (SemEval-2020 Task 7) - `dev/core.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English), Yelp Review Full - `dev/probes.jsonl`: unspecified: TweetEval offensive (OLID) - `dev/sources/sentiment/amazon-reviews.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `dev/sources/sentiment/imdb.jsonl`: unspecified: Large Movie Review Dataset (IMDB) - `dev/sources/sentiment/sst5.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `dev/sources/sentiment/yelp-full.jsonl`: restricted: Yelp Review Full - `dev/sources/topic/trec.jsonl`: unspecified: TREC question classification - `dev/sources/transfer/tweeteval-offensive.jsonl`: unspecified: TweetEval offensive (OLID) - `dev/transfer.jsonl`: unspecified: TweetEval offensive (OLID) - `sources/eval/breadth/arts/cfcolor/cfcolor.zip`: unspecified: Collaborative Filtering of Color Aesthetics - `sources/eval/breadth/arts/humicroedit/semeval-2020-task-7-dataset.zip`: unspecified: Humicroedit (SemEval-2020 Task 7) - `sources/eval/breadth/language/hellaswag/data/validation-00000-of-00001.parquet`: restricted: HellaSwag validation file (raw, with wikiHow items) - `sources/eval/breadth/tools/routerbench/routerbench_0shot.pkl`: restricted: RouterBench 0-shot file (raw) - `sources/eval/transfer/tweeteval-offensive/offensive/test-00000-of-00001.parquet`: unspecified: TweetEval offensive (OLID) - `sources/eval/transfer/tweeteval-offensive/offensive/train-00000-of-00001.parquet`: unspecified: TweetEval offensive (OLID) - `sources/eval/transfer/tweeteval-offensive/offensive/validation-00000-of-00001.parquet`: unspecified: TweetEval offensive (OLID) - `sources/train/sentiment/amazon-reviews/test.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `sources/train/sentiment/amazon-reviews/train.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `sources/train/sentiment/amazon-reviews/validation.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `sources/train/sentiment/imdb/plain_text/test-00000-of-00001.parquet`: unspecified: Large Movie Review Dataset (IMDB) - `sources/train/sentiment/imdb/plain_text/train-00000-of-00001.parquet`: unspecified: Large Movie Review Dataset (IMDB) - `sources/train/sentiment/sst5/dev.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `sources/train/sentiment/sst5/test.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `sources/train/sentiment/sst5/train.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `sources/train/sentiment/yelp-full/yelp_review_full/test-00000-of-00001.parquet`: restricted: Yelp Review Full - `sources/train/sentiment/yelp-full/yelp_review_full/train-00000-of-00001.parquet`: restricted: Yelp Review Full - `sources/train/topic/trec/default/test/0000.parquet`: unspecified: TREC question classification - `sources/train/topic/trec/default/train/0000.parquet`: unspecified: TREC question classification - `test/breadth.jsonl`: unspecified: Collaborative Filtering of Color Aesthetics, Humicroedit (SemEval-2020 Task 7) - `test/core.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English), Yelp Review Full - `test/probes.jsonl`: unspecified: TweetEval offensive (OLID) - `test/sources/sentiment/amazon-reviews.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `test/sources/sentiment/imdb.jsonl`: unspecified: Large Movie Review Dataset (IMDB) - `test/sources/sentiment/sst5.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `test/sources/sentiment/yelp-full.jsonl`: restricted: Yelp Review Full - `test/sources/topic/trec.jsonl`: unspecified: TREC question classification - `test/sources/transfer/tweeteval-offensive.jsonl`: unspecified: TweetEval offensive (OLID) - `test/transfer.jsonl`: unspecified: TweetEval offensive (OLID) - `train/core.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English), Yelp Review Full - `train/sources/sentiment/amazon-reviews.jsonl`: restricted: Multilingual Amazon Reviews Corpus (English) - `train/sources/sentiment/imdb.jsonl`: unspecified: Large Movie Review Dataset (IMDB) - `train/sources/sentiment/sst5.jsonl`: unspecified: Stanford Sentiment Treebank (SST-5) - `train/sources/sentiment/yelp-full.jsonl`: restricted: Yelp Review Full - `train/sources/topic/trec.jsonl`: unspecified: TREC question classification Classes are conservative readings of each source's own terms, recorded with their evidence; they are not legal advice.