Adapts to changing defenses
We have not verified a test showing a response to changed defenses. Avoiding obstacles is a different capability.
Last reviewed
What we are looking for
Does the system notice a change in a defense and change its behavior?
Where this applies
No qualifying system/version in the publications we checked
What has been observed
- A 2026 preprint reports generalization to new racetracks without test-time adaptation.Green et al. / arXiv · Jun 25, 2026
What limits this finding
- The paper identifies changed appearance as a failure risk for Swift.UZH / Nature · Aug 30, 2023
What we still need to know
A defense-change experiment separating onboard response from operator intervention.
What would change our view
A primary test report explicitly meeting this criterion.
What would be enough to confirm it
An observed response to a changed defense in a physical test; obstacle avoidance and offline retraining are distinct.