# 研究揭示机器人在线学习遗忘机制：接管式纠正可精确移除用户影响

> 原文链接：https://robotics.hellobike.com/papers/2026-11-10-2611-10280-14d823c7
> 频道：papers | 更新日期：2026-11-10 | 标签：auto, arxiv
> 核心摘要：针对机器人部署后在线学习中特定用户数据移除难题，研究证实仅当用户通过接管控制进行纠正时，按原始时序重训剩余会话可精确恢复目标策略；单纯删除数据重训因丢失时序依赖而失效，动作微调类纠正则导致影响扩散无法完全消除。

# 研究揭示机器人在线学习遗忘机制：接管式纠正可精确移除用户影响

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[Who Taught the Robot That? Removing One User's Corrections from a Robot Policy That Learns Online](https://arxiv.org/abs/2611.10280)（arXiv 2611.10280）

## 核心内容

arXiv 2611.10280 论文探讨了机器人在多人连续在线纠正场景下执行“机器遗忘”的可行性边界。研究在两个模拟任务中设定脚本化用户，以“若该用户从未存在过的理想策略”为基准进行量化评估。核心发现指出，遗忘效果严格取决于人类纠正交互的模态：只有当用户采用全权接管控制方式时，系统才能通过保留有序更新日志并按时序重训剩余数据，达到与训练噪声水平相当的精确移除效果。

技术方案层面，研究揭示了在线学习对数据时序的高度敏感性。对于接管式纠正，直接删除目标用户数据并进行单次重训会破坏训练更新的因果顺序，即使后续用户纠正不显式依赖被删数据也无法收敛至目标策略。相反，若用户仅对机器人动作进行微调（Nudge），其影响会通过其他用户的后续纠正渗透进模型参数及训练缓冲区，导致现有测试方法均无法恢复目标策略；此时继续训练虽能消除大部分习惯，但无法清除缓冲区内残留的数据依赖。

## 关键看点

- **接管纠正可逆性**：仅当用户通过接管控制纠正时，利用有序更新日志按原始时序重训剩余会话，可精确逼近无该用户时的理想策略状态。
- **时序依赖陷阱**：直接删除用户数据并重训因丢失训练更新顺序而失效，证明在线学习策略对数据到达时序具有不可忽略的因果依赖性。
- **微调影响扩散**：动作微调类纠正会导致被删用户影响渗透至其他用户数据及训练缓冲区，现有方法无法完全剥离此类隐式耦合影响。
- **影响力评估局限**：训练日志仅能识别明确有害用户，但在量化用户影响力排序方面，复杂算法并未显著优于简单的纠正次数统计。

## 工程与产业观察

该研究为具身智能在线学习与合规数据删除提供了关键工程约束：在设计人机共学系统时，应优先采用“接管式”而非“微调式”交互接口以保障策略的可逆性与可审计性。同时，底层训练框架必须维护带时间戳的原子级更新日志，而非仅存储数据集快照，这是实现符合GDPR等法规要求的精确机器遗忘、避免模型污染的必要基础设施条件。

## 英文摘要 (Abstract)

A robot that keeps learning after deployment is corrected by several people in turn, and one of them may later ask to be removed. We show that whether the robot can forget that person depends on how people correct it. The goal of removal is the robot that would exist if the person had never been there, which we compute exactly in two simulated tasks with scripted users. When users correct the robot by taking over control, an ordered log of training updates is enough: retraining on the remaining sessions in their original order gives a robot as close to the goal as training noise allows. Deleting the person's data and retraining in one pass does not, even when the other users' corrections do not depend on the removed person, because it loses the order of training. When users only nudge the robot's action, the other users' corrections carry the removed person's influence, and no method we tested recovers the goal robot. Continuing to train without the person removes most of their habits, but not what their data in the training buffer keep alive. The log can identify one clearly harmful user, but not rank users by influence better than counting their corrections.

作者：Eshika Pathak, Leela Krishna
:::

## 原始参考来源 (Sources)

- [arXiv 2611.10280：Who Taught the Robot That? Removing One User's Corrections from a Robot Policy That Learns Online](https://arxiv.org/abs/2611.10280)