| language: | |
| - de | |
| task_categories: | |
| - text-generation | |
| size_categories: | |
| - 100k<n<1M | |
| task_ids: | |
| - language-modeling | |
| - masked-language-modeling | |
| German validation dataset from WECHSEL () to evaluate LLM perplexity. | |
| JSON-line files (on JSON object per line): | |
| - `valid.json.gz`: Gzipped validation set as generated by the paper (163,698 docs) | |
| - `valid.random_1636.json.gz`: Random 1% (1636 docs) of the validation set | |