Certified Behavioral Retention for Continual Adaptation with Held-Out Anchors

Continual adaptation requires deciding not only how to update a model, but whether to commit an update without exceeding a declared budget for changes to protected behavior.

持续适应不仅需要决定如何更新模型，还需要决定是否在不超过受保护行为变更声明预算的情况下提交更新。

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 largest tested interpolation coefficient whose worst-anchor forward KL divergence remains within budget.

我们引入了基于锚点的认证适应（CABA），这是一种与候选模型无关的更新后门控机制。它在候选优化过程中保留历史观测数据，在冻结的表征上比较旧策略与候选策略，并选择最大的测试插值系数，以确保最差锚点的前向 KL 散度保持在预算范围内。

The finite check requires neither action labels nor complete trajectories. We develop a theory-led certification framework that separates measured anchor retention from conditional off-anchor guarantees.

这种有限检查既不需要动作标签，也不需要完整的轨迹。我们开发了一个理论导向的认证框架，将测量的锚点保留与条件化的非锚点保证分离开来。

Under explicit representation-coverage and Lipschitz assumptions, an accepted update satisfies a uniform behavioral-divergence bound on protected support.

在明确的表征覆盖和 Lipschitz 假设下，被接受的更新满足受保护支持集上的统一行为散度界限。

The analysis gives regularity bounds for affine softmax heads, probabilistic coverage and compression guarantees, interpolation monotonicity conditions, grid-resolution bounds, and trajectory-level and sequential-composition guarantees.

该分析给出了仿射 Softmax 头的正则性界限、概率覆盖和压缩保证、插值单调性条件、网格分辨率界限，以及轨迹级和顺序组合的保证。

In a controlled UCI Human Activity Recognition study used as a non-stationary sensor-policy proxy, CABA reduces mean forgetting from 1.01 percent for unconstrained adaptation to 0.00 percent at budget epsilon = 0.05, while reducing final average accuracy from 93.04 percent to 88.48 percent.

在一项用作非平稳传感器策略代理的受控 UCI 人类活动识别研究中，当预算 epsilon = 0.05 时，CABA 将平均遗忘率从无约束适应的 1.01% 降低至 0.00%，同时将最终平均准确率从 93.04% 降低至 88.48%。

The experiment illustrates a retention-plasticity trade-off and does not establish foundation-model or robotic performance.

该实验说明了保留与可塑性之间的权衡，并未确立基础模型或机器人性能。

CABA provides an auditable retention check for candidate generators exposing comparable policy distributions and a declared update path, while deployment also requires separate evidence of current-task utility.

CABA 为暴露可比策略分布和声明更新路径的候选生成器提供了一种可审计的保留检查，而部署该机制还需要当前任务效用的独立证据。