argilla/mistral-vs-llama

该数据集包含一个符合Argilla数据集格式的配置文件`argilla.yaml`，以及兼容HuggingFace `datasets`库的记录。数据集可以用于不同的NLP任务，具体取决于配置。数据集中包含字段、问题、建议和指南。字段是数据集记录本身，目前仅支持文本字段。问题是向注释者提出的问题，可以是评分、文本、单选或多选类型。建议是提供给注释者的辅助信息，以简化注释过程。数据集结构部分详细描述了数据字段和问题类型。

原始信息汇总

数据集卡片 for mistral-vs-llama

数据集描述

数据集概述

该数据集包含：

• 符合 Argilla 数据集格式的配置文件 argilla.yaml。该配置文件将在使用 Argilla 的 FeedbackDataset.from_huggingface 方法时用于配置数据集。
• 与 HuggingFace datasets 兼容的数据集记录。这些记录在使用 FeedbackDataset.from_huggingface 时会自动加载，也可以通过 datasets 库独立加载。
• 用于构建和整理数据集的标注指南（如果已在 Argilla 中定义）。

加载方式

使用 Argilla 加载

安装 Argilla：

bash pip install argilla --upgrade

加载数据集：

python import argilla as rg

ds = rg.FeedbackDataset.from_huggingface("argilla/mistral-vs-llama")

使用 datasets 加载

安装 datasets：

bash pip install datasets --upgrade

加载数据集：

python from datasets import load_dataset

ds = load_dataset("argilla/mistral-vs-llama")

支持的任务和排行榜

该数据集可以包含多个字段、问题和响应，因此可以用于不同的 NLP 任务，具体取决于配置。数据集结构在数据集结构部分中描述。

该数据集没有关联的排行榜。

语言

[更多信息需要]

数据集结构

数据在 Argilla 中

数据集在 Argilla 中包含以下内容：字段、问题、建议和指南。

字段

字段名称	标题	类型	必需	Markdown
input	输入	TextField	True	False
response_1	响应_1	TextField	True	True
response_2	响应_2	TextField	True	True

问题

问题名称	标题	类型	必需	描述	值/标签
ranking	对响应进行排序	RankingQuestion	True	N/A	[response_1, response_2]
rationale_1	响应_1排序的理由？	TextQuestion	True	N/A	N/A
rationale_2	响应_2排序的理由？	TextQuestion	True	N/A	N/A

建议

✨ NEW 此外，我们还有建议，这些建议与现有问题相关联，并且包含建议的值及其元数据。

指南

指南是一个纯字符串，用于向标注者提供指令。请参见标注指南部分。

数据实例

Argilla 中的数据实例

json { "fields": { "input": "How do I locate my card?", "response_1": " Thank you for reaching out! Iu0027m here to help you with your query. However, I must point out that the term "locate" can have different meanings depending on the context. Could you please provide more details or clarify what you mean by "locate"? Are you trying to find a specific card, such as a credit card or debit card? Or perhaps youu0027re looking for something else entirely? Once I have a better understanding of your question, Iu0027ll be happy to assist you in a responsible and accurate manner.", "response_2": " Of course! Iu0027d be happy to help you locate your missing card. To start, could you please tell me more about the card youu0027re trying to find? Is it a credit card, debit card, or another type of card? Additionally, can you provide more details about the circumstances surrounding its disappearance? For example, did you lose it at home, in a store, or somewhere else? Any additional information you can provide will help me give you more tailored advice on how to locate your card." }, "metadata": {}, "responses": [ { "status": "submitted", "user_id": "39bdc103-0e37-4515-a353-abe3b68bdebd", "values": { "ranking": { "value": [ { "rank": 2, "value": "response_1" }, { "rank": 1, "value": "response_2" } ] }, "rationale_1": { "value": "The text is correct and provides a helpful response to the query. However, it lacks specific instructions or steps to help the user locate their card, hence itu0027s not highly informative." }, "rationale_2": { "value": "The text is accurate and provides a more detailed response to the query. It asks for more specific information to provide a tailored solution, making it highly informative. However, it still lacks the profound insights or comprehensive information that would make it outstandingly helpful." } } } ], "suggestions": [ { "question_id": "9820a347-3573-46cf-88e3-c172e02c6f96", "question_name": "ranking", "value": [ { "rank": 1, "value": "response_2" }, { "rank": 2, "value": "response_1" } ] }, { "agent": "gpt-4", "question_id": "177a00ed-8536-4cd5-a755-d4436cfd1455", "question_name": "rationale_1", "value": "The text is correct and provides a helpful response to the query. However, it lacks specific instructions or steps to help the user locate their card, hence itu0027s not highly informative." }, { "agent": "gpt-4", "question_id": "7bd8451e-4c63-491b-afd8-8aabb67ce62f", "question_name": "rationale_2", "value": "The text is accurate and provides a more detailed response to the query. It asks for more specific information to provide a tailored solution, making it highly informative. However, it still lacks the profound insights or comprehensive information that would make it outstandingly helpful." } ] }

HuggingFace datasets 中的数据实例

json { "external_id": null, "input": "How do I locate my card?", "metadata": "{}", "ranking": [ { "status": "submitted", "user_id": "39bdc103-0e37-4515-a353-abe3b68bdebd", "value": { "rank": [ 2, 1 ], "value": [ "response_1", "response_2" ] } } ], "ranking-suggestion": { "rank": [ 1, 2 ], "value": [ "response_2", "response_1" ] }, "ranking-suggestion-metadata": { "agent": null, "score": null, "type": null }, "rationale_1": [ { "status": "submitted", "user_id": "39bdc103-0e37-4515-a353-abe3b68bdebd", "value": "The text is correct and provides a helpful response to the query. However, it lacks specific instructions or steps to help the user locate their card, hence itu0027s not highly informative." } ], "rationale_1-suggestion": "The text is correct and provides a helpful response to the query. However, it lacks specific instructions or steps to help the user locate their card, hence itu0027s not highly informative.", "rationale_1-suggestion-metadata": { "agent": "gpt-4", "score": null, "type": null }, "rationale_2": [ { "status": "submitted", "user_id": "39bdc103-0e37-4515-a353-abe3b68bdebd", "value": "The text is accurate and provides a more detailed response to the query. It asks for more specific information to provide a tailored solution, making it highly informative. However, it still lacks the profound insights or comprehensive information that would make it outstandingly helpful." } ], "rationale_2-suggestion": "The text is accurate and provides a more detailed response to the query. It asks for more specific information to provide a tailored solution, making it highly informative. However, it still lacks the profound insights or comprehensive information that would make it outstandingly helpful.", "rationale_2-suggestion-metadata": { "agent": "gpt-4", "score": null, "type": null }, "response_1": " Thank you for reaching out! Iu0027m here to help you with your query. However, I must point out that the term "locate" can have different meanings depending on the context. Could you please provide more details or clarify what you mean by "locate"? Are you trying to find a specific card, such as a credit card or debit card? Or perhaps youu0027re looking for something else entirely? Once I have a better understanding of your question, Iu0027ll be happy to assist you in a responsible and accurate manner.", "response_2": " Of course! Iu0027d be happy to help you locate your missing card. To start, could you please tell me more about the card youu0027re trying to find? Is it a credit card, debit card, or another type of card? Additionally, can you provide more details about the circumstances surrounding its disappearance? For example, did you lose it at home, in a store, or somewhere else? Any additional information you can provide will help me give you more tailored advice on how to locate your card." }

数据字段

数据集字段包括以下内容：

• 字段：这些是数据集记录本身，目前仅支持文本字段。这些字段将用于提供问题的响应。

  • input 是 TextField 类型。
  • response_1 是 TextField 类型。
  • response_2 是 TextField 类型。

• 问题：这些问题将向标注者提出。它们可以是不同类型，如 RatingQuestion、TextQuestion、LabelQuestion、MultiLabelQuestion 和 RankingQuestion。

  • ranking 是 RankingQuestion 类型，允许值为 [response_1, response_2]。
  • rationale_1 是 TextQuestion 类型。
  • rationale_2 是 TextQuestion 类型。

• ✨ NEW 建议：从 Argilla 1.13.0 开始，建议已包含在内，以向标注者提供建议，以简化或协助标注过程。建议与现有问题相关联，总是可选的，并且不仅包含建议本身，还包含其相关元数据（如果适用）。

  • (可选) ranking-suggestion 是 ranking 类型，允许值为 [response_1, response_2]。
  • (可选) rationale_1-suggestion 是 text 类型。
  • (可选) rationale_2-suggestion 是 text 类型。

此外，还有一个可选字段：

• external_id：这是一个可选字段，可用于为数据集记录提供外部 ID。如果您希望将数据集记录链接到外部资源（如数据库或文件），这可能很有用。

数据分割

数据集包含一个分割，即 train。