OopsieRecover: Learning Transferable VLA Corrections from Semantic Interventions
OopsieRecover: Learning Transferable VLA Corrections from Semantic Interventions
Vision-Language-Action (VLA) models often fail during deployment. OopsieRecover is a framework that uses abduction to infer the intent behind corrective demonstrations and encodes this knowledge as dense spatial representations and conditional costs. Experiments across six tasks show that only a few corrective demonstrations enable VLA models to overcome failures in novel situations without requiring new data for each variation.
The authors propose OopsieRecover, a framework that uses abduction to infer intent from semantic interventions and encode corrections as dense spatial representations and conditional costs, enabling VLA models to generalize failure recovery across novel objects and scenes with only a handful of demonstrations.
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.
- code: not_reported
- weights: not_reported
- dataset: not_reported
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
tasks, and find that only a handful of corrective demonstrations can enable a VLA to overcome failures across a broad range of novel situations, without requiring new corrective data for each variation. Please refer to https://robin-lab.cs.utexas.edu/oopsierecover for videos.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org