sk-bench
Collection
26 items • Updated
You need to agree to share your Hugging Face information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
Access Request Terms:
By requesting access to this dataset, you confirm that you:
- will cite the sk-bench paper and any additional works listed in the Citation section of this dataset card, and respect the dataset's license;
- will ensure that your downstream use complies with applicable laws, regulations, and ethical AI principles.
Log in or Sign Up to review the conditions and access this dataset content.
Dataset Card for chiby-classification
Dataset Details
Dataset Description
This is a Slovak dataset intended for the classification of sentences as either grammatically correct or incorrect. It is a preprocessed and annotated version of dataset Chiby v0.4, which is based on revisions of Slovak Wikipedia. This dataset contains 3432 rows.
The dataset was cleaned by removing HTML tags from sentences, and errors were extracted into new columns for annotation purposes. Sentence similarity was then verified using cosine similarity to ensure correct pairing between the original sentence and the revised sentence. From seven revision types in the original dataset, three were selected: spelling, diacritics, and capitalization. All sentence pairs containing more than one type of error were removed. After preprocessing, 67% of pairs contained spelling errors, 23% diacritics errors, and 10% capitalization errors.
The dataset was subsequently annotated (see the Annotation section for more details).
A second version of this dataset, called chiby-error-correction, is intended for error correction. In this version, one column contains the incorrect sentence and the second column contains the corrected sentence.
Dataset Structure
Each instance contains:
{
"sentence":"V depozitári bolo umiestnených asi 800 predmetov, prevažne obrazy a nábytok z 18. a 19. storočia.",
"label":1
}
Fields
| Field | Type | Description |
|---|---|---|
sentence |
string | Sentence in Slovak language |
label |
int64 | 0 = incorrect sentence, 1 = correct sentence |
Annotations
Annotation process
After preprocessing, 2400 sentence pairs were randomly selected: 240 from the capitalisation category, 552 from the diacritics category, and 1608 from the spelling category, matching the distribution of the preprocessed dataset. Each sentence pair was annotated by a single annotator. Either the first sentence or the second sentence was incorrect, as the original dataset, Chiby v0.4, relies on automatic revision. In some cases, both sentences were incorrect. When the grammatical error was minor, it was corrected and the pair was retained in the dataset. After annotation, 1716 sentence pairs remained, and were randomly shuffled. The sentences were labelled as 0 if incorrect, and 1 if correct, resulting in a total of 3432 rows in the dataset.
Who are the annotators?
The dataset was annotated by three native speakers, all of whom have a university-level education.
Citation
If you use the dataset chiby-classification, please cite the following:
@inproceedings{suppa-etal-2026-sk,
title = "sk-bench: A Native-First Benchmark for Evaluating Large Language Models in {S}lovak",
author = "{\v{S}}uppa, Marek and
Vykopal, Ivan and
Ridzik, Andrej and
Sopkovi{\v{c}}, Kristi{\'a}n and
K{\v{n}}a{\v{z}}ekov{\'a}, Nat{\'a}lia and
Kopcan, Jaroslav and
Blstak, Miroslav and
Ondrejov{\'a}, Vikt{\'o}ria and
Hl{\'a}dek, Daniel and
Gregor, Michal and
Tamajka, Martin and
Simko, Marian",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Junyi Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2639/",
doi = "10.18653/2026.emnlp-main.2639",
pages = "60008--60055",
ISBN = "979-8-89176-459-0",
}
@article{Garabik2026,
author = {Radovan Garabík},
title = {Slovenská Wikipédia ako zdroj dát pre korpus chýb},
journal = {Kultúra slova},
year = {2026},
note = {In print}
}
License
This dataset is released under the Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0) license, the same as the Chiby v0.4 dataset.
- Downloads last month
- 21