onetimetwo/toxic-chat
收藏资源简介:
该数据集包含从Vicuna在线演示收集的10K用户提示的毒性标注。采用人机协作标注框架确保标注质量,同时保持可行的标注工作量。数据集包含两个版本(toxicchat0124和toxicchat1123),最新版本0124修正了之前版本的标注错误并增加了额外标注样本。数据集可用于训练和评估毒性检测模型,促进构建安全健康的用户-AI交互环境。
This dataset contains toxicity annotations on 10K user prompts collected from the Vicuna online demo. It utilizes a human-AI collaborative annotation framework to guarantee the quality of annotation while maintaining a feasible annotation workload. There are two versions (toxicchat0124 and toxicchat1123), with the latest version 0124 fixing annotation errors from the previous version and adding additional annotated samples. The dataset can be used to train and evaluate toxicity detection models, promoting advancements toward building a safe and healthy environment for user-AI interactions.




