unlearning-cleanslate/fsid-curated-olmo-32b-target-100

该数据集包含多个配置（forget、forget_pool、retain、retain_pool），用于研究语言模型的记忆和遗忘行为。特征包括内容ID、标题、歌词、记忆分数（memorized_fraction）、窗口索引等，涉及文本生成和评估指标（如ROUGE-L、BLEU）。数据集可能用于分析模型对特定内容（如歌曲歌词）的记忆程度，并包含不同分割（如baseline、bm25_10B）以比较不同条件。

This dataset includes multiple configurations (forget, forget_pool, retain, retain_pool) for studying memory and forgetting behaviors in language models. Features include content ID, title, lyrics, memorized fraction, window index, etc., covering text generation and evaluation metrics (e.g., ROUGE-L, BLEU). It is likely used to analyze the degree of model memorization for specific content (e.g., song lyrics) and contains different splits (e.g., baseline, bm25_10B) for comparison under various conditions.