Chelsea Finn：机器人成功率越高，DAgger 效率越低

DAgger 在机器人高成功率下效率低

AI 导读

Chelsea Finn 指出，随着机器人越来越可靠，DAgger 的效率会大幅下降，例如 90% 的成功率意味着人超过 90% 的时间都在被动看着机器人。她提出的核心思路是：失败来自相似的初始状态，可以把学习聚焦在这些状态上。相关论文与 Lars Ankile 的 Mulligan 工作一致，后者在 3 个真实任务、2550 次盲测中，在相同预算下优于均匀采集。

正文 · 原文

Chelsea Finn· @chelseabfinn · X·2026-10-10 05:38· 22 分钟前AI 评分34

AI 导读

随着机器人变得越来越可靠，DAgger 的效率就大大降低了。例如 90% 的成功率意味着人 >90% 的时间都在被动地看着机器人核心思路：失败来自相似的初始状态，我们可以把学习聚焦在这些状态上！论文：https://arxiv.org/abs/2610.05882

正文

As robots become more reliable, DAgger becomes far less efficient.

eg 90% success means the person spends >90% of the time passively watching the robot

Key idea: failures come from similar initial states & we can focus learning on these states!

Paper: https://arxiv.org/abs/2610.05882

引用Lars Ankile@larsankile

I've been thinking a lot about making robot learning scalable. To me, that means learning on the job, and using the human help still needed efficiently. In Mulligan ⛳️, we focus each data-collection round on the initial states where the robot fails. This happens in deployment, with the robot doing the task and a person stepping in when needed. On 3 real tasks and 2,550 blind evals, it improves on uniform collection at the same budget.

在 X 查看被引用的帖子

来源：Chelsea Finn · x.com

来源：RadarAI · aihot.news