Data-Gouv-FR/compar-ia
收藏资源简介:
Compar:IA是一个对话式AI比较器(聊天机器人竞技场),由法国文化部开发,具有双重使命:培训和提升对模型多样性、文化语言偏见及对话式AI环境问题的认识;收集和发布用于观察使用情况和改进法语对话式AI的问题和偏好数据集。该比较器在compar:IA初创公司(隶属于法国数字局DINUM的beta.gouv.fr计划)框架下开发,帮助公共机构构建有用、简单易用的数字服务。收集的数据发布为三个数据集:1. Conversations compar:IA:包含平台上所有提问和回答的数据集。2. Votes compar:IA:汇总用户在完整对话后表达偏好的数据集,用于在没有单个消息反应时比较两个模型。3. Réactions compar:IA:汇总用户在消息级别反应的数据集,反映在讨论过程中逐条消息表达的偏好。
Compar:IA is a conversational AI comparator (a chatbot arena) developed within the French Ministry of Culture with a dual mission: to train and raise awareness about model diversity, cultural and linguistic biases, and environmental issues of conversational AIs; and to create and publish datasets of questions and preferences useful for observing usage and improving conversational AIs in French. The comparator is developed as part of the compar:IA startup (integrated into the beta.gouv.fr program of the Interministerial Directorate for Digital Affairs, DINUM), which helps public administrations build useful, simple, and easy-to-use digital services. The data collected via compar:IA is published in three datasets: 1. Conversations compar:IA: A dataset of all questions asked and answers received on the compar:IA platform. 2. Votes compar:IA: A dataset aggregating user preferences expressed after an entire conversation, used to compare two models when no reactions have been given on individual messages. 3. Réactions compar:IA: A dataset aggregating user reactions at the message level, reflecting preferences expressed message by message during discussions.




