ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning
ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning
This research proposes ActionUNet, an efficient multi-scale fine-tuning framework for Vision-Language-Action (VLA) models. By integrating a lightweight temporal U-Net and a conditional SIREN decoder, the method enhances temporal continuity and reduces mechanical oscillations, leading to improved success rates in simulated and real-world robotic tasks.
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
Reported success rate
9.8%
View original evidence
execution failures while preserving the base VLA model's generalization and manipulation robustness. Extensive experiments on RoboTwin 2.0 and LIBERO-Plus benchmarks, together with real-world hard evaluations, demonstrate that ActionUNet significantly improves {\pi}0.5 success rates by absolute 9.8%, 6.1%, and 11.4%, respectively, while also generalizing to the regression-based OpenVLA-OFT backbone, highlighting its
Open source S5
- code: Available Artifact link Open source S6
- code: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
execution failures while preserving the base VLA model's generalization and manipulation robustness. Extensive experiments on RoboTwin 2.0 and LIBERO-Plus benchmarks, together with real-world hard evaluations, demonstrate that ActionUNet significantly improves {\pi}0.5 success rates by absolute 9.8%, 6.1%, and 11.4%, respectively, while also generalizing to the regression-based OpenVLA-OFT backbone, highlighting its
Open source S5
effectiveness and efficiency as a fine-tuning strategy. Code and implementation details are available at https://github.com/Di-Zhu123/ActionUNet.
Open source S6
Source：arXiv Robotics — research abstracts · arxiv.org