felipeadachi/chatbot_arena_harmful_annotated

该数据集源自lmsys的Chatbot Arena Conversations Dataset，包含用户提示（user_prompt）和表示是否有害意图的二进制标签（is_harmful）。数据集还包含了用于生成is_harmful标签的LLM模型信息。在标注过程中，有135行数据没有被正确处理，保留了值为-的记录。数据集的目的是根据提供的用户提示，将其分类为无有害意图或有害意图。

This dataset is derived from lmsys Chatbot Arena Conversations Dataset and includes user prompts and a binary label indicating harmful intent. It also contains information about the LLM model used to generate the harmful intent labels. During the annotation process, 135 rows were not correctly processed and retained with a value of -. The dataset is intended to classify user prompts into non-harmful or harmful intention based on the provided prompts.