提出新方法让大模型持续编辑不崩溃，稳定提升长期编辑效果。
Norm Anchors Make Model Edits Last
- 通过归一化锚点重设向量幅度，打破编辑过程中的正反馈失控循环。
- 在多个模型和数据集上将可编辑轮次提升4倍以上，长期性能平均提高72.2%。
- 仅需一行代码修改，计算开销极低，适合所有现有编辑工具使用。
连续定位与编辑（L&E）模型编辑在多次操作后会突然失效。我们识别并形式化这一现象为正范数反馈环：修复后的向量与更新的MLP权重相互放大，导致编辑质量下降并最终破坏模型能力。分析表明，在标准L&E动态下，该反馈可引发近指数级的范数增长，且现有增量正则化或更新钳制无法解决。为此，我们提出范数锚点缩放（NAS），一种即插即用的稳定器，通过将每个修复向量缩放到原始模型参考范数来打破此循环。在多个LLM骨干、数据集和编辑器上，NAS将可用编辑周期扩展超过4倍，平均提升长期编辑性能72.2%，同时保持单次编辑效果，仅需一行代码修改且计算开销可忽略。
原文摘要 · Abstract (English)
Sequential Locate-and-Edit (L&E) model editing can fail abruptly after many edits. We identify and formalize this failure as a positive norm-feedback loop, in which solved value vectors and edited MLP weights progressively amplify each other, degrading edit quality and eventually collapsing model capabilities. Our analysis shows that this feedback can yield approximately exponential norm growth under standard L&E dynamics, and can remain unresolved by existing increment-level regularizers or update clamps. We propose Norm-Anchor Scaling (NAS), a plug-in stabilizer that breaks this loop by rescaling each solved value vector to an original-model reference norm. Across multiple LLM backbones, datasets, and L&E editors, NAS extends the usable editing horizon by more than 4x and improves long-run editing performance by 72.2% on average, while preserving single-edit efficacy, with only a one-line modification and negligible computational overhead.
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。