Update README.md

README.md

@@ -18,7 +18,36 @@ dataset_info:
     num_examples: 627
   download_size: 52557
   dataset_size: 91941
+license: cc
+language:
+- cs
+pretty_name: Czech grammar agreement dataset
 ---
-# Dataset Card for "agree-cs"
+# Czech grammar agreement dataset (AGREE)
+
+This is an adapted an filtered test subset from the original [Czech grammar agreement dataset](https://nlp.fi.muni.cz/~xbaisa/agree/), 
+designed to evaluate Czech language competence in the subject-verb agreement problem.
+
+The examples were transformed to accommodate a missing word selection task. 
+Sentences containing more than one marked verb were discarded. 
+In the remaining sentences, the marked verb is completely replaced with the "____" token. 
+All five possible verb variants form the list of available choices. The index of the correct choice is stored as the label.
+Preblamatic examples were identified by gradually selecting examples wrongly answered by Claude 3 Haiku, Claude 3 Sonet and GPT-4 Turbo. These 115 examples were then manually checked and 46 of them were identified as ambiguous and removed from the dataset. This led to a final count of 627 evaluation samples.
+
+This dataset was created for use within the [Czech-Bench](https://gitlab.com/jirkoada/czech-bench) evaluation framework. 
+
+## Citation
+
+```bibtex
+@PhdThesis{Baisa2016thesis,
+AUTHOR = "Baisa, Vít",
+TITLE = "Byte Level Language Models [online]",
+YEAR = "2016 [cit. 2024-08-28]",
+TYPE = "Disertační práce",
+SCHOOL = "Masarykova univerzita, Fakulta informatiky, Brno",
+NOTE = "SUPERVISOR : Karel Pala",
+URL = "https://is.muni.cz/th/en6ay/",
+}
+```
+
 
-[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)