aisi-whitebox/wmdp_cyber_cot_prompted_sandbagging_llama_31_8b_instruct_finetuning

该数据集包含聊天对话内容及相关信息，每个样本包含对话内容(content)、角色(role)、目标(targets)、得分(scores)、答案(answers)、系统提示(sys_prompts)等字段。数据集还包含样本是否为良性(is_benign)、输入ID(input_ids)、任务名称(task_name)、样本索引(sample_index)、数据集ID(dataset_id)、sandbagging执行情况(sandbagging_executed)和提示的数据集名称(prompted_dataset_name)等信息。数据集分为训练集(train)等部分，训练集共有311个样本。

The dataset contains chat conversation content and related information, with each sample including fields such as content, role, targets, scores, answers, sys_prompts, etc. The dataset also includes information such as whether the sample is benign (is_benign), input ID (input_ids), task name (task_name), sample index (sample_index), dataset ID (dataset_id), execution of sandbagging (sandbagging_executed), and the name of the dataset prompted (prompted_dataset_name). The dataset is split into sections such as training set (train), with a total of 311 samples in the training set.