AirsideLabs/A-CDM

该数据集包含了多种配置的数据，每种配置具有不同的特征字段。主要包含以下几类数据： 1. chunked配置：包含文档ID、文档文本、文档文件名、文档元数据（如文件大小）、原始摘要、摘要、文档摘要、摘要模型、文本块信息、多跳文本块信息、文本块信息指标和分块模型等字段。 2. ingested配置：包含与chunked配置类似的字段，但缺少了文本块和多跳文本块信息。 3. lighteval配置：包含问题、额外说明、真实答案、答案类型、难度评估、引用等信息。 4. multi_hop_questions配置：包含多跳问题的相关信息，如文档ID、源文本块ID、问题、自我答案等。 5. single_shot_questions配置：包含单次问答的相关信息。 6. summarized配置：与chunked配置类似，但专注于文档摘要信息。The dataset includes multiple configurations with different feature fields. The main types of data are as follows: 1. chunked configuration: includes fields such as document ID, document text, document filename, document metadata (e.g., file size), raw summaries, summaries, document summary, summarization model, text chunk information, multi-hop text chunk information, text chunk metric information, and chunking model. 2. ingested configuration: includes similar fields to the chunked configuration but lacks text chunk and multi-hop text chunk information. 3. lighteval configuration: includes fields such as question, additional instructions, ground truth answer, answer type, difficulty assessment, citations, etc. 4. multi_hop_questions configuration: includes information related to multi-hop questions, such as document ID, source text chunk IDs, question, self-answer, etc. 5. single_shot_questions configuration: includes information related to single-shot questions. 6. summarized configuration: similar to the chunked configuration but focuses on document summary information.