--- dataset_info: - config_name: collection features: - name: title dtype: string - name: text dtype: string - name: passage_id dtype: int64 splits: - name: train num_bytes: 3915958652 num_examples: 2308679 download_size: 2190188831 dataset_size: 3915958652 - config_name: generated features: - name: passage_id dtype: int64 - name: title dtype: string - name: text dtype: string - name: output dtype: string - name: model dtype: string splits: - name: train num_bytes: 4664331696.887888 num_examples: 1697195 download_size: 2334314297 dataset_size: 4664331696.887888 configs: - config_name: collection data_files: - split: train path: collection/train-* - config_name: generated data_files: - split: train path: generated/train-* license: cc-by-sa-4.0 language: - ja --- 日本語Wikipediaから抽出したテキストに基づいて、[rinna/deepseek-r1-distill-qwen2.5-bakeneko-32b](https://huggingface.co/rinna/deepseek-r1-distill-qwen2.5-bakeneko-32b)と[https://huggingface.co/cyberagent/DeepSeek-R1-Distill-Qwen-32B-Japanese](https://huggingface.co/cyberagent/DeepSeek-R1-Distill-Qwen-32B-Japanese)を用いて箇条書きにしたデータセットです。 重複を許してランダムにサンプリングしながら生成をしたので、`collection` subsetにある事例が`generated` subsetに無い場合があります。 また、HF上のviewerでは改行や空白がうまく見えませんが、実際にはほとんどの箇条書きの要素が改行で区切られています。 生成に用いたモデルがApache 2.0で配布されていること、日本語Wikipediaのテキストを用いていることから、本データセットはCC-BY-SA 4.0での配布とします。