unlearning-cleanslate/eval-llama-3_1-8b-simnpo-gentle-bm25-10b
收藏数据链接:
官方服务:
资源简介:
该数据集包含用于评估语言模型记忆的文本片段及其相关统计信息,包括文本长度、滑动窗口分析结果(如记忆窗口数、覆盖率、p_z值等)、最佳记忆窗口的详细信息(种子、目标、位置等),以及评估模型参数和内容元数据。
This dataset contains text fragments and related statistics for evaluating language model memorization, including text length, sliding window analysis results (e.g., memorized windows, coverage, p_z values), detailed information of the best memorized window (seed, target, positions), evaluation model parameters, and content metadata.



