Discussly builds on open datasets published by researchers and volunteers. We are grateful to every one of them. Questions have been filtered, de-duplicated and edited for clarity; any errors introduced by our edits are ours.
Datasets used
- BoolQ (Google Research) — licensed under CC BY-SA 3.0.
- OpenTriviaQA — licensed under CC BY-SA 4.0.
- Open Trivia Database — licensed under CC BY-SA 4.0.
- CommonsenseQA — licensed under MIT.
- AI2 Reasoning Challenge — licensed under CC BY-SA 4.0.
- MMLU (Hendrycks et al.) — licensed under MIT.
- GSM8K (OpenAI) — licensed under MIT.
- args.me corpus — licensed under CC BY 4.0.
- TruthfulQA — licensed under Apache 2.0.
- IBM Debater — Argument Quality 30k — licensed under CC BY 3.0.
Reusing our question sets
Each collection page lists the datasets its questions come from. You may reuse a collection under the license of its sources: for CC BY and CC BY-SA sources, credit the original dataset (and Discussly) and, for CC BY-SA, share your adapted version under the same license. Collections from MIT or Apache 2.0 sources may be reused with the license notice kept.
The Discussly website design, code and the how-to guides on each page are not covered by these licenses.