What Happens to the Goal? Distinct Failures of Policy Alignment in Goal Misgeneralization
What Happens to the Goal? Distinct Failures of Policy Alignment in Goal Misgeneralization
This research explores how agents fail to maintain goal alignment during distribution shift, identifying three distinct failure modes based on where the goal is lost within the agent's internal processing. The study shows that recovery methods vary depending on the depth of goal loss and demonstrates how steering and downstream wiring can improve goal reach in different scenarios.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
task
8
View original evidence
etraining. We study this across eight visual navigation and manipulation tasks, including robot control. While tracing where goal information remains available internally, how strongly it influences the policy, and what role it plays in action selection, we uncover that the goal is lost at one of three different depths. This depth predicts which recovery works in each task. When the goal is followed, but outcompeted
Open source S3
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
by the proxy, steering the existing policy redirects it to the goal's location, raising goal reach from 16.0% to 95.3%. When it is read, but not followed, steering fails, but a lightweight new downstream wiring recovers goal reach from 31.6% to 84.3%. When it is not read, none of the methods work, as the policy does not use the goal information. The same tracing also reveals what replaces the goal: functional target
Open source S4
etraining. We study this across eight visual navigation and manipulation tasks, including robot control. While tracing where goal information remains available internally, how strongly it influences the policy, and what role it plays in action selection, we uncover that the goal is lost at one of three different depths. This depth predicts which recovery works in each task. When the goal is followed, but outcompeted
Open source S3
substitution, where behavior follows an alternative goal, a policy shortcut, where behavior follows a learned action strategy, or cue competition between intended and proxy goal cues. Knowing where the goal is lost tells us whether an agent can be repaired or has to be retrained. Goal misgeneralization should therefore be studied and evaluated inside the agent, not only from its behavior.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org