PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations
PDRVLA: A Vision-Language-Action Model with Physical Dynamics Representations
This paper presents PDRVLA, a new Vision-Language-Action (VLA) model that integrates a lightweight Physical Dynamics Module to extract physical tokens from vision-language features for conditioning action generation.
The study introduces PDRVLA, a Vision-Language-Action model that incorporates a lightweight Physical Dynamics Module to condition action generation on predicted future RGB, depth, and optical-flow features without online video generation, trained entirely on publicly available data augmented by human-to-robot synthesis and embodied chain-of-thought annotations.
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.
- Environment
- SimulationOpen source S4
Reported numbers
Reported success rate
99.3%
Reported trials
View original evidence
hought (ECoT) annotation. Across four simulation benchmarks, PDRVLA achieves success rates of 99.3% on LIBERO, 89.9% on LIBERO-Plus, 93.6% on RoboTwin~2.0, and 81.5% on SimplerEnv. Ablations show improved future prediction with world-model supervision and evaluate how subtask conditioning, future object-pose supervision, and ECoT training affect action success.
Open source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
What remains unknown
Not established in the collected evidence: 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.
hought (ECoT) annotation. Across four simulation benchmarks, PDRVLA achieves success rates of 99.3% on LIBERO, 89.9% on LIBERO-Plus, 93.6% on RoboTwin~2.0, and 81.5% on SimplerEnv. Ablations show improved future prediction with world-model supervision and evaluate how subtask conditioning, future object-pose supervision, and ECoT training affect action success.
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org