REDIRECT: A 1% Fix for Bad Robot Habits

What happened

arXiv:2610.03997v1 Announce Type: new Abstract: Robots can acquire (one company buying another) bad habits from a few defective moments in otherwise useful teleoperation (a person controlling a machine from somewhere else). Across three randomized ManiSkill tasks and three seeds, REDIRECT raises the mean success rate from 68.2% to 90.3%, recovering 88.8% of the clean-retraining gap versus 22.1% for matched-compute fine-tuning (further training of an existing model for a narrower job).

In mixed-quality robot data, normal and problematic demonstrations share most task behavior and differ only at a local action continuation. Full retraining is costly, while fine-tuning on clean data alone offers limited recovery under a small update budget.

Using only episode-level retained/problematic labels, REDIRECT localizes the branch, assigns a coherent retained continuation to the original observations around it, and anchors shared behavior, without frame-level annotations or additional interaction. On a PiPER arm, it recovers 86.7-92.9% of the clean-only gap across cup insertion and towel folding.