schneiderkamplab/dala-cs-audited

DaLA Czech — audited grammar and spelling 393,609 synthetic original/corrupted pairs, selected by an automated Gemma audit. Not native-speaker validated or a gold benchmark. No paraphrasing, simplification or style-transfer task. Valid variants should not be labelled errors; model judgments remain fallible and systematic errors can survive filtering. Split Pairs Rows per task train 315,029 630,058 validation 39,163 78,326 test 39,417 78,834 Two… See the full description on the dataset page: https://huggingface.co/datasets/schneiderkamplab/dala-cs-audited.

<iframe src="https://silytics.si/embed/item/bbcc3dd2-d8f9-4273-a42a-62bbb7bc1823" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.