identity_dpo
收藏魔搭社区2025-12-05 更新2025-10-11 收录
下载链接：
https://modelscope.cn/datasets/mrfakename/identity_dpo
下载链接
链接失效反馈官方服务：
资源简介：
# Identity DPO v0 (Alpha)
> Teaching LLMs not to think they're ChatGPT
A small DPO (Direct Performance Optimization) dataset for teaching LLMs not to think they're ChatGPT.
## Dataset Creation
This data was created through a mix of human and synthetic data and was reviewed and curated by human reviewers.
## Notes
* No synthetic data was generated by ChatGPT. All synthetic data was generated through properly-licensed LLMs.
# 身份DPO v0（Alpha测试版）
> 调教大语言模型（Large Language Model）切勿将自身误认为ChatGPT
本数据集为轻量级直接性能优化（Direct Performance Optimization，DPO）数据集，旨在引导大语言模型切勿将自身误认为ChatGPT。
## 数据集构建
本数据集通过人工数据与合成数据混合构建，并经人工评审人员审核与筛选整理。
## 注意事项
* 所有合成数据均非由ChatGPT生成，全部合成数据均通过获得合法授权的大语言模型生成。
提供机构：
maas
创建时间：
2025-07-24
