基于留出锚点的持续适应行为保持认证
原标题：Certified Behavioral Retention for Continual Adaptation with Held-Out Anchors
AI 摘要
论文提出 Certified Anchor-Based Adaptation (CABA)，一种与候选生成器无关的更新后门控机制：在候选优化前预留历史观测作为锚点，在冻结表示上比较新旧策略，并选择满足最坏锚点前向 KL 不超过预设预算的最大插值系数。作者给出理论认证框架，将实测锚点保持与锚点外条件保证分离，并在 UCI 人体活动识别数据集上验证：预算 ε=0.05 时平均遗忘从 1.01 个百分点降至 0.00，但最终平均准确率从 93.04% 降至 88.48%。
正文节选
Certified Behavioral Retention for Continual Adaptation with Held-Out Anchors Abstract Continual adaptation requires deciding not only how to update a model, but whether to commit the update without exceeding a declared budget for changes to protected behavior. We introduce Certified Anchor-Based Adaptation (CABA), a candidate-agnostic post-update gate that reserves historical observations from candidate optimization, compares old and candidate policies on frozen representations, and selects the