Update README.md
Browse files
README.md
CHANGED
@@ -765,7 +765,52 @@ configs:
|
|
765 |
data_files:
|
766 |
- split: train
|
767 |
path: data/train-*
|
|
|
|
|
|
|
|
|
|
|
|
|
768 |
---
|
769 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
770 |
|
771 |
-
[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
|
|
765 |
data_files:
|
766 |
- split: train
|
767 |
path: data/train-*
|
768 |
+
license: cc-by-4.0
|
769 |
+
language:
|
770 |
+
- da
|
771 |
+
pretty_name: Danoliterate Survey Answers
|
772 |
+
size_categories:
|
773 |
+
- n<1K
|
774 |
---
|
775 |
+
# Danoliterate Survey Answers
|
776 |
+
- Updated: 2024 08 28
|
777 |
+
|
778 |
+
## Introduction
|
779 |
+
- Results of the Danoliterate Survey with Human Feedback on Generative, Large Language Models in Danish.
|
780 |
+
|
781 |
+
- If you speak, Danish add your own response: [danoliterate.compute.dtu.dk/Spørgeskema](https://danoliterate.compute.dtu.dk/Spørgeskema)
|
782 |
+
|
783 |
+
- For more details, see [danoliterate.compute.dtu.dk/Articles](https://danoliterate.compute.dtu.dk/Articles#results-of-the-danoliterate-survey)
|
784 |
+
|
785 |
+
- The data was produced by the survey implementation in the [Danoliterate Frontend](https://github.com/sorenmulli/danoliterate-frontend/) hosted at [danoliterate.compute.dtu.dk](https://danoliterate.compute.dtu.dk/)
|
786 |
+
|
787 |
+
- The prompts used for the evaluation are available in [sorenmulli/danoliterate-survey-prompts](https://huggingface.co/datasets/sorenmulli/danoliterate-survey-prompts)
|
788 |
+
|
789 |
+
-
|
790 |
+
## Data format
|
791 |
+
Each row is one A/B test carried out by a human.
|
792 |
+
No preprocessing has been carried out from the frontend except translating the emojis to a score from 0 to 4 and removing the text response column.
|
793 |
+
Furthermore, a bug before May 23 means that the seen prompts feature is null for these.
|
794 |
+
|
795 |
+
### Fields
|
796 |
+
|
797 |
+
- `model_A` and `model_B`: The two models compared in this test.
|
798 |
+
|
799 |
+
- `prefer`: The chosen model between A, B, "Ved ikke" or no answer.
|
800 |
+
|
801 |
+
- `likert-A` and `likert-B`: The score optionally given to the models from 0 to 4.
|
802 |
+
|
803 |
+
- `seen_prompts`: The prompts shown to the user before the test.
|
804 |
+
|
805 |
+
- `index`: The index of this test within the session.
|
806 |
+
|
807 |
+
- `was_revealed`: Whether the the user after answering chose to see the models.
|
808 |
+
|
809 |
+
- `user-gender`, `user-age`, `user-language`, `user-experience`: The user's self-reported demographics.
|
810 |
+
|
811 |
+
- `session-id`: ID of the session that this test is part of.
|
812 |
+
|
813 |
+
- `session-all-chosen-models`, `session-all-was-revelead`, `session-all-seen-prompts`: The aggregated data of the session.
|
814 |
+
|
815 |
+
- `session-timestamp`: The timestamp of the session.
|
816 |
|
|