RynnVLA-Lat, Scaling Robotic Manipulation Foundation Models with 3D-Grounded Latent Actions
RynnVLA-Latent: Scaling Robotic Manipulation Foundation Models with 3D-Grounded Latent Actions
This research presents RynnVLA-Latent, a framework that enables scalable pretraining of robotic manipulation policies by learning 3D-grounded latent actions from heterogeneous interaction videos. The method avoids reliance on native robot controls and leverages visual transitions, future-feature prediction, and camera-motion decoupling to capture control-relevant motion. The results show strong performance on multiple benchmarks, including LIBERO, LIBERO-Plus, VLABench, and RoboTwin 2.0, with success rates exceeding those of competing methods in several real-world 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
latent-action descriptors
745,520,000
Trajectory count
View original evidence
al transitions without requiring native robot controls. By combining residual future-feature prediction, camera-compensated 3D-flow supervision, and camera-motion decoupling, RynnLAM encourages representations that capture control-relevant motion rather than appearance changes alone. We apply RynnLAM to 30.1K hours of human, robot, and simulation videos, producing 745.52M latent-action descriptors. RynnVLA-Latent the
Open source S3
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
al transitions without requiring native robot controls. By combining residual future-feature prediction, camera-compensated 3D-flow supervision, and camera-motion decoupling, RynnLAM encourages representations that capture control-relevant motion rather than appearance changes alone. We apply RynnLAM to 30.1K hours of human, robot, and simulation videos, producing 745.52M latent-action descriptors. RynnVLA-Latent the
Open source S3
n learns to predict latent-action chunks from visual observations and language through flow matching, before adapting its pretrained backbone and action expert to embodiment-specific robot actions. The resulting 4B policy achieves 98.8% success on LIBERO, 84.7% on LIBERO-Plus without benchmark-specific adaptation, 51.9% average success across five VLABench tracks, and 88.6% average success on the RoboTwin 2.0 bimanua
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org