agentlans/en-chat-refusal
收藏Hugging Face2025-12-19 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/agentlans/en-chat-refusal
下载链接
链接失效反馈官方服务：
资源简介：
这是一个包含50万英语对话的数据集，从大型数据库中采样并使用NousResearch/Minos-v1拒绝分类器进行标注。数据集包含平衡的拒绝和非拒绝标签数据行。输入字段是使用Snowflake/snowflake-arctic-embed-xs分词器进行分词后的对话摘要版本，消息之间用[SEP]标记分隔。数据集存在一定的局限性，包括分类器可能出现的误判情况（假阳性和假阴性），以及一些边界案例。
500,000 English conversations sampled from a large database and annotated using NousResearch/Minos-v1 refusal classifier. The refusal_data split contains balanced data with both refusal and non-refusal labelled rows. The input field is an abridged version of the conversation as tokenized using the Snowflake/snowflake-arctic-embed-xs tokenizer, with messages separated by [SEP] tokens. The dataset has known limitations including potential classifier errors (false positives and negatives) and edge cases.
提供机构：
agentlans
