英文标题：What a Deletion Certificate Covers, and Where It Expires: Auditable Removal from a Support-Vector Memory
作者：Vishwajith Ramesh
arXiv ID：2607.12204 | 分类：cs.LG | 发表：2026-09-11
许可：CC-BY
摘要 注意力可被视为一种基于上下文的在线学习器，然而现有的测试时记忆无法认证丢弃一个token后输出保持不变，也无法彻底删除其影响。我们提出支撑向量注意力（SV-Attention），一种最大间隔记忆，其权重是固定箱参数{{PT_MATH_1}}的单类SVM的支撑系数。其活跃集划分赋予保留token精确的零权重，从而认证保持输出的驱逐；一个可逆的增量求解器删除一个token，以恢复在相同{{PT_MATH_2}}下重新训练（不含该token）所产生的状态。在fp64实验中，当最优解唯一时，递减与重新拟合恢复出相同的划分，且它们的决策函数中位数偏差约为{{PT_MATH_3}}（在学习到的键上为
We study deletion in a context memory that fits a support-vector boundary around stored keys and uses the resulting nonnegative coefficients to weight their values. An exactly zero coefficient lets us remove a key without changing the current normalized readout. In a three-key construction, however, a later admission makes the discarded key active in a solve over the full history. For a positive-weight key, a decremental update targets a refit on the remaining keys at the original coefficient cap, under its stated assumptions. Earlier standalone comparisons show close typical gate-score agreement, while similar keys with different values can yield larger readout differences. The journal extension tests chained deletions and admissions, checking coefficient feasibility and optimality conditions in both returned states and fresh references. The initial audit completes 66 of 80 trajectories and exposes numerical failures. A separately checked wrapper completes all 5,120 scheduled operations at unchanged acceptance thresholds, with eight initialization rescues and 84 update rescues. Its largest readout difference is 0.5273% of retained-value range. These finite results support explicit acceptance and rescue rules for the evaluated memory edits. They do not extend the current-state removal guarantee to arbitrary future admissions.
查看完整双语翻译 →
正在跳转到翻译阅读页… 如果没有自动跳转，请点击这里。