保护模型免被非法合并，保持原模型性能同时大幅降低合并后效果
Defending Unauthorized Model Merging via Dual-Stage Weight Protection
- 分两阶段保护：分散关键参数并注入结构扰动
- 合并后模型准确率下降最高达90%，原模型损失低于1.5%
- 适用于视觉与语言模型，防止未经授权的模型融合
预训练模型的快速传播和开源库的普及使得模型合并变得便捷但风险增加，自由骑乘者可未经许可将微调模型合并为多功能新模型，侵犯知识产权并破坏所有权。为此，我们提出MergeGuard，一种主动的双阶段权重保护框架，在保持任务性能的同时破坏合并兼容性。第一阶段通过L2正则化优化，将任务相关信息均匀分布于各层；第二阶段注入结构扰动，使任务子空间错位，破坏损失曲面中的曲率兼容性。两者协同重塑模型参数几何结构，导致合并模型产生破坏性干扰，而受保护模型仍保持完整功能。在视觉（ViT-L-14）和语言模型（Llama2、Gemma2、Mistral）上的大量实验表明，MergeGuard可使合并模型准确率下降最高达90%，且受保护模型性能损失小于1.5%。
原文摘要 · Abstract (English)
The rapid proliferation of pretrained models and open repositories has made model merging a convenient yet risky practice, allowing free-riders to combine fine-tuned models into a new multi-capability model without authorization. Such unauthorized model merging not only violates intellectual property rights but also undermines model ownership and accountability. To address this issue, we present MergeGuard, a proactive dual-stage weight protection framework that disrupts merging compatibility while maintaining task fidelity. In the first stage, we redistribute task-relevant information across layers via L2-regularized optimization, ensuring that important gradients are evenly dispersed. In the second stage, we inject structured perturbations to misalign task subspaces, breaking curvature compatibility in the loss landscape. Together, these stages reshape the model's parameter geometry such that merged models collapse into destructive interference while the protected model remains fully functional. Extensive experiments on both vision (ViT-L-14) and language (Llama2, Gemma2, Mistral) models demonstrate that MergeGuard reduces merged model accuracy by up to 90% with less than 1.5% performance loss on the protected model.
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。