akseljoonas/toxic-chat-filtered
收藏Hugging Face2025-12-17 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/akseljoonas/toxic-chat-filtered
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含对话数据，用于评估对话模型的安全性和合规性。主要特征包括对话ID、用户输入、模型输出、人工标注、毒性评分、越狱尝试和OpenAI审核结果。数据集分为训练集(6998样本)、验证集(1500样本)和测试集(1500样本)三部分。
This dataset contains conversational data for evaluating the safety and compliance of dialogue models. Key features include conversation ID, user input, model output, human annotation, toxicity score, jailbreaking attempts, and OpenAI moderation results. The dataset is divided into train (6,998 examples), validation (1,500 examples), and test (1,500 examples) sets.
提供机构：
akseljoonas
