inLoRA-MINT：用中间表征审计 LoRA 微调模型的训练数据
inLoRA-MINT: Internal Representations for Auditing Training Data in Domain-adapted LLMs
能判断某个样本有没有被拿去 LoRA 微调，在 Qwen、LLaMA 等四个模型上准确率超 95%。
论文提出 inLoRA-MINT，把成员推断测试 MINT 应用到 LoRA 微调的大语言模型上，判断某个样本是否参与过训练。方法利用激活值等中间表征训练 MINT 分类器，无需完整访问模型参数，区别于此前仅靠困惑度的 LoRA-MINT 方案。研究团队在 Qwen、Phi-4、DeepSeek-V3 和 LLaMA 3.2-3B 四个模型上，用医疗、数学、学术推理三类数据集做微调测试。区分训练样本与非训练样本的成绩稳定在 95% 到 98%，对模型配置和数据分布变化保持稳健。作者称该方法不限于 LoRA，可推广到其他模型适配技术，用于数据溯源和隐私合规审计。
inLoRA-MINT: Internal Representations for Auditing Training Data in Domain-adapted LLMs
This work investigates the application of the Membership Inference Test (MINT) to large language models (LLMs) fine-tuned with LoRA, aiming to determine whether specific samples were used during training and address concerns about data privacy, leakage, and transparency. Building on our previous LoRA-MINT framework, which relies exclusively on perplexity under a strict black-box setting, we propose an approach based on intermediate model representations, including activations and parameter-dependent signals. These representations enable a MINT classifier to distinguish training from non-training samples without requiring full access to model parameters. We evaluated the method on four LLMs (Qwen, Phi-4, DeepSeek-V3, and LLaMA 3.2-3B) fine-tuned on three datasets covering medical, mathematical, and academic reasoning tasks. Results show consistent performance across architectures and domains, achieving scores between 95% and 98%, with robustness to variations in model configurations and data distributions. The proposed approach enables efficient auditing of sensitive training data and supports data provenance and compliance with emerging privacy and AI governance regulations. Although our experiments focus on LoRA-based fine-tuning, the methodology can be extended to other LLM adaptation techniques and, more broadly, to other AI models.
