PACT-VLA: Speculative Action-Chunk Verification with Pre-Commit Observations
PACT-VLA: Speculative Action-Chunk Verification with Pre-Commit Observations
This paper presents PACT-VLA, a training-free framework that improves the success rate of vision-language-action (VLA) policies by using pre-commit observations to select the best action chunk during inference delay. It outperforms existing methods on both simulated and real-robot tasks, achieving significant improvements in task success rates.
PACT-VLA introduces a novel approach to address inference delay in vision-language-action policies by leveraging pre-commit observations for candidate selection, improving task success rates in both simulated and real-world settings.
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
delay steps
4
View original evidence
ity predicted by the frozen policy under the pre-commit observation, requiring no separate verifier. A pipelined implementation overlaps observation processing with candidate generation and batches scoring, so selection adds no control-step delay on the evaluated hardware. At a four-step delay, PACT-VLA improves average success by 15.0 percentage points over RTC on Kinetix and by 5.2 points over single-candidate asyn
Open source S5
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ity predicted by the frozen policy under the pre-commit observation, requiring no separate verifier. A pipelined implementation overlaps observation processing with candidate generation and batches scoring, so selection adds no control-step delay on the evaluated hardware. At a four-step delay, PACT-VLA improves average success by 15.0 percentage points over RTC on Kinetix and by 5.2 points over single-candidate asyn
Open source S5
chronous execution with GR00T N1.7 on LIBERO, and it further improves existing delay-robust methods. On three real-robot tasks, it raises average success from 60% to 82% over naive asynchronous execution with the same frozen $\pi_{0.5}$ checkpoint.
Open source S6
Source：arXiv Robotics — research abstracts · arxiv.org