Lakera/gandalf-rct-did
收藏官方服务:
资源简介:
该数据集包含了多个字段,包括提示(prompt)、答案(answer)、原始答案(raw_answer)、llm、设置(setup)、防御(defense)、难度等级(level)和试验ID(trial_id)。数据集分为一个子集defense_in_depth_experiment,其中包含了超过十万个例子。该数据集主要用于深度防御实验。
The dataset includes multiple fields such as prompt, answer, raw_answer, llm, setup, defense, level, and trial_id. The dataset is split into one subset called defense_in_depth_experiment, which contains over one hundred thousand examples. This dataset is primarily used for in-depth defense experiments.
提供机构:
Lakera


