repnoise_beavertail
收藏资源简介:
该数据集是一个对话数据集,包含了对话的发起内容(prompt)和响应内容(response)。每个对话都被标记了是否包含以下类别的话题:虐待动物、虐待儿童、有争议的话题(包括政治)、歧视、刻板印象、不公正、滥用药物、武器、禁用物质、金融犯罪、财产犯罪、盗窃、仇恨言论、攻击性语言、关于伦理、法律和安全的错误信息、非暴力不道德行为、侵犯隐私、自残、成人内容(包括露骨的性内容)、恐怖主义、有组织犯罪、暴力、教唆和煽动。此外,还有一个标签用于指示内容是否安全,以及一个拒绝标签,其内容为字符串类型。数据集分为训练集,共有1000个示例,数据集大小为1,692,687字节。
This dataset is a conversational dataset containing the initiation content (prompt) and response content (response) of dialogues. Each dialogue is labeled whether it covers topics from the following categories: animal abuse, child abuse, controversial topics (including politics), discrimination, stereotypes, injustice, substance abuse, weapons, banned substances, financial crimes, property crimes, theft, hate speech, offensive language, misinformation regarding ethics, law and safety, non-violent unethical conduct, privacy violations, self-harm, adult content (including explicit sexual content), terrorism, organized crime, violence, incitement and instigation. Additionally, there is a label indicating whether the content is safe, as well as a rejection label with a string value. The dataset is split into a training set which includes 1000 examples, with a total size of 1,692,687 bytes.




