This is a benchmark for Sweden-related factual knowledge. Its questions are inspired by the hosts of the Swedish radio program Sommar I P1 as well as sports-related events in Sweden (e.g., events that are part of En Svensk Klassiker.
Answers are designed to be minimal to enable simple string-based answer recall to approximate model performance as well as possible. Note that some samples in the dataset have multiple correct answers; if so, they are separated by commas.
See the preprint for… See the full description on the dataset page:
https://huggingface.co/datasets/liu-nlp/swedish-facts-v1.
