提出JUMP方法，高效检测微调语言模型是否使用了特定数据。
JUMP: Single-Pass Membership Inference on Fine-Tuned Diffusion Language Models
- 基于参考模型选择低置信度位置联合掩码，一次推理完成检测。
- 在六个数据集上将检测性能提升至0.902~0.942的ROC-AUC。
- 仅需3次模型前向传播，比现有方法快10倍以上，适合实际部署。
公开的开源语言模型常在私有或领域特定数据上进行微调后部署，因此需要审计其中是否使用了特定记录。本文研究离散扩散语言模型（dLLMs）的成员身份推断问题，以微调前的检查点作为参考。与自回归模型不同，dLLMs可并行处理任意掩码集合，并同时输出所有被掩码位置的预测结果。现有方法SAMA通过多次随机掩码平均重建信号，但会稀释关键位置信息，且需重复多次模型评估。本文提出JUMP（联合不确定性引导掩码探测），选择参考模型置信度低的位置进行联合掩码，每样本仅用一次评分查询，聚合截断后的目标-参考重建误差差值。在六个MIMIR领域中，JUMP使LLaDA-8B-Base的平均ROC-AUC从0.819提升至0.902，Dream-v0-Base-7B从0.851提升至0.942，而模型前向传播次数从SAMA的32次降至3次。
原文摘要 · Abstract (English)
Public open-weight language models are often fine-tuned on private or domain-specific data before deployment, creating a need to audit whether individual records were used during adaptation. We study this problem for discrete diffusion language models (dLLMs), using the pre-fine-tuning checkpoint as a reference. Unlike autoregressive models, dLLMs allow arbitrary mask sets and return predictions for all masked positions in parallel. SAMA averages reconstruction signals over many random masks, which can dilute informative positions and requires repeated model evaluations. We propose JUMP (Joint Uncertainty-Guided Mask Probing), which selects low-reference-confidence positions, masks them jointly, and aggregates clipped target-reference reconstruction gaps from one scoring query per model. Across six MIMIR domains, JUMP raises mean ROC-AUC from 0.819 to 0.902 on LLaDA-8B-Base and from 0.851 to 0.942 on Dream-v0-Base-7B, while using three model forwards per sample versus 32 for SAMA.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。