CASE EVIDENCE / A RECORD
Google Research / google-research-datasets 公开的代码代理与软件工程案例，来源为 公开代码库，复核于 2026-06-27T09:46:00Z。
原始记录：Google Research used Gemini Ultra to generate the Education Dialogue Dataset
Chinese Brief
Google Research / google-research-datasets 公开的代码代理与软件工程案例，来源为 公开代码库，复核于 2026-06-27T09:46:00Z。 Model Atlas 将它标记为 A 类证据，因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。 Model Atlas 不把 benchmark、教程、发布说明或集合页包装成真实案例。
任务
这是一个围绕软件工程任务执行的真实任务，公开材料可以回溯到具体使用者和具体产物。 原始资料写作：Google Research published the Education Dialogue Dataset; its README states that the conversations were generated by prompting Gemini Ultra, with a teacher persona teaching a specified topic and a student persona respon…
公开材料提供公开代码、README 或项目配置，可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作：The public GitHub repository contains a released education-dialogue dataset with train and test JSON files, including multi-turn teacher/student conversations plus metadata such as topic, student preferences, teacher pr…
Gemini 1.0 Ultra 在该案例中承担软件工程任务执行相关的生成、分析、编排或实现角色。 原始资料写作：Gemini 1.0 Ultra, referenced by the project as Gemini Ultra, generated the core synthetic teacher-student dialogue content from the dataset prompts, making the model the production source for the released conversation a…
当前判断基于公开材料；若产物下线、仓库变更或模型参与比例仅来自作者自述，需要在引用前重新复核。 原始资料写作：The README names Gemini Ultra rather than spelling out 'Gemini 1.0 Ultra'; because this public dataset was released in the Gemini 1.0 Ultra era and the task allows model-family binding, it is mapped to gemini-1-0-ultra.…