aisi-whitebox/wmdp_bio_cot_prompted_sandbagging_llama_31_8b_instruct_finetuning_finetuning

该数据集包含了聊天对话内容、角色信息、目标、元数据、得分、答案、系统提示、是否良性、输入ID、任务名称、样本索引、数据集ID、sandbagging执行情况、提示的数据集名称和分割信息等字段。其中，聊天对话内容包含对话内容和对话角色信息。数据集的元数据字段为null类型，得分和答案字段为字符串类型。此外，数据集还提供了训练集的分割信息，包括字节数和示例数量。

The dataset includes fields such as chat content, role information, targets, metadata, scores, answers, system prompts, benign or not, input IDs, task names, sample indices, dataset IDs, execution of sandbagging, prompted dataset names, and split information. The chat content includes conversation content and the roles in the conversation. The metadata field in the dataset is of type null, and the score and answer fields are of string type. In addition, the dataset provides split information for the training set, including the number of bytes and the number of examples.