研究人员开发了一种新的参数高效微调方法，该方法将适应性限制在从现有任务适配器派生的子空间内。此方法旨在通过限制可达到的更新来缓解微调投毒。在具有 196 个 LoRA 适配器的 FLAN-T5-Large 上进行的实验表明，这种子空间约束的适应性可以在干净数据上匹配完整的 LoRA 性能，同时显著提高对标签反转攻击和后门尝试的抵抗力。 AI
影响 这项研究可以增强微调模型免受恶意攻击的安全性，使其在下游应用中更加可靠。
排序理由 该集群包含一篇详细介绍 AI 模型微调新方法的论文。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2607.05300v1 Announce Type: new Abstract: Parameter-efficient fine-tuning still leaves a broad space of behavior-changing updates reachable, so a poisoned objective can be represented and optimized. We study an alternative: adaptation constrained to the subspace estimated f…
Parameter-efficient fine-tuning still leaves a broad space of behavior-changing updates reachable, so a poisoned objective can be represented and optimized. We study an alternative: adaptation constrained to the subspace estimated from a trusted pool of existing task adapters. On…
Parameter-efficient fine-tuning still leaves a broad space of behavior-changing updates reachable, so a poisoned objective can be represented and optimized. We study an alternative: adaptation constrained to the subspace estimated from a trusted pool of existing task adapters. On…