CIIRC-NLP/intention_recognition_guardrail_dataset
收藏Hugging Face2025-12-15 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/CIIRC-NLP/intention_recognition_guardrail_dataset
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含用于训练Amazon Nova AI Challenge 2025的Itention Recognition模型的数据，目的是训练一个护栏系统，用于判断对编码LLM的请求是否安全。数据包含三种标签：良性（benign）、边界线（borderline）和恶意（malicious）。数据中还包含多轮样本，但仅包含用户提示而没有响应。如果历史中有恶意提示，则会被占位符文本替换。
This dataset contains data used for training of the Itention Recognition model for the Amazon Nova AI Challenge 2025 for the AlquistCoder team. The purpose of this dataset was to train a guardrail which determines whether a request to a coding LLM is safe to answer. The data contains three labels: benign, borderline, and malicious. Multiturn samples are also present, however only the user prompts without responses. If a malicious prompt would be in history, it is instead replaced by a placeholder text (this message would be blocked).
提供机构：
CIIRC-NLP
