DuongTrongChi/DTX-Detection

该数据集包含多个特征，主要涉及指令、越南语指令、拒绝分数、提示分数和预测。指令特征包含角色和文本两个子特征。拒绝分数和提示分数都是结构体，包含多个毒性相关的分数，如亵渎、严重毒性、侮辱、威胁、毒性和身份攻击。数据集包含一个训练集，共有9995个样本，总大小为3209511.267450404字节。

The dataset contains multiple features, including instruction, instruction_vi (Vietnamese instruction), score_rejected, score_prompt, and predict. The instruction feature is a list containing two sub-features: role and text. Both score_rejected and score_prompt are structures containing multiple toxicity-related scores, such as PROFANITY, SEVERE_TOXICITY, INSULT, THREAT, TOXICITY, and IDENTITY_ATTACK. The dataset includes a training set with 9995 examples and a total size of 3209511.267450404 bytes.