详解LLM预训练数据中的叙事篇章特征，揭示模型如何理解和生成故事结构，对改进训练数据质量有重要启示
arXiv:2606.19468v1 Announce Type: new Abstract: The narrative composition of web-scale LLM pretraining corpora remains largely unexplored even though …
聚焦日本LLM预训练数据中的敏感个人信息检测，为模型训练隐私安全提供新思路
arXiv:2606.12114v1 Announce Type: new Abstract: Sensitive personal information can appear in large-scale pre-training corpora for large language model…
揭秘LLM「说谎」根源：论文提出用检索预训练数据替代纯参数知识，实现更诚实的AI输出。
arXiv:2601.21218v2 Announce Type: replace Abstract: Large language models (LLMs) are highly capable of answering questions, but they are often unaware…
提出黑盒场景下掩码语料级预训练数据检测新方法，助力LLM数据泄露评估
arXiv:2606.07996v1 Announce Type: new Abstract: Pretraining is fundamental to the development of Large Language Models (LLMs), yet the opacity of pret…
提出TANDEM双层优化框架，利用孪生网络自动搜索最优数据配比，为预训练数据混合提供新解法。
arXiv:2606.04401v1 Announce Type: new Abstract: The capabilities of large language models (LLMs) significantly depend on training data drawn from vari…
提出基于top-1预测差距的Gap-K%方法，精准检测LLM预训练数据，破解隐私与版权难题
arXiv:2601.19936v2 Announce Type: replace-cross Abstract: The opacity of massive pretraining corpora in Large Language Models (LLMs) raises significan…
系统梳理LLM预训练数据暴露的成员推理、数据污染与安全影响，已被NLDB 2025接收，是研究数据隐私的必备综述。
arXiv:2605.26133v1 Announce Type: cross Abstract: Large Language Models (LLMs) have become the predominant paradigm in NLP, advancing both research an…