Xense Robotics Helps Anyverse Dynamics MWA™-Phoenix Top Six Leaderboards
Xense Robotics Helps Anyverse Dynamics MWA™-Phoenix Top Six Leaderboards
Anyverse Dynamics has open-sourced MWA™-Phoenix, a 7-billion-parameter world model that topped six simulation and real-robot leaderboards, including LIBERO-Plus and RoboDojo. The model uses tactile data from Xense Robotics sensors and a Phoenix-ILA module to improve success rates in tasks like cucumber peeling and plate-wiping. Real-world interaction data enhances action learning, and the model's performance is validated through simulation and real-robot testing.
The paper reports the open-sourcing of MWA™-Phoenix, a 7-billion-parameter world model that achieved top rankings on six simulation and real-robot leaderboards. It highlights the integration of tactile data from Xense Robotics sensors and the use of a Phoenix-ILA module for joint learning of motion and tactile signals, resulting in improved success rates for manipulation tasks.
Source: Robotics — Robot learning news discovery · Read original article ↗
Article text · Original source · English
Gasgoo Munich- Anyverse Dynamics has released and open-sourced MWA™-Phoenix, a fully pre-trained, 7-billion-parameter latent-space world model foundation built for edge deployment.
Public benchmarks show the model topping six simulation and real-robot leaderboards — including LIBERO-Plus, LIBERO-Pro, and the RoboDojo real-robot ranking. It hit a 94.63% overall success rate on LIBERO-Plus. Across all real-robot demonstrations released this time, Xense Robotics's photon tactile sensors served as the tactile input source, feeding real contact data directly into action learning and interaction prediction.
Image source: Xense Robotics Robotics
On the action-learning front, Anyverse Dynamics' Phoenix-ILA module encodes synchronized motion and tactile sequences into a shared interactive latent space, jointly learning movement, local tactile deformation, and force-related signals. Ablation studies reveal that adding this module doubled the success rate for cucumber peeling — from 20% to 40% — and lifted plate-wiping accuracy from 75% to 90%. For tasks demanding continuous surface contact, real-world interaction data gives action learning a direct grounding.
During pre-deployment policy training, Anyverse Dynamics' in-house Phoenix-PhysCanvas visual-tactile world model predicts the visual and tactile feedback after an action is executed, supporting reinforcement-learning post-training. Training runs in an interactive simulation environment before results are validated on real robots. Experiments show average real-robot success rates climbing from 58.33% to 75.00% for test-tube tasks, and from 80.00% to 86.67% for block-stacking tasks — gains of 16.67 and 6.67 percentage points, respectively.
The real-robot tasks — twisting bottle caps, peeling, wiping, and long-horizon operations — all rely on Xense Robotics sensors for signal capture. To safeguard data quality, Xense Robotics built cross-unit consistency and repeatability testing standards into its sensor R&D and manufacturing, strictly aligning the temporal relationships among visual, tactile, and action signals. Meanwhile, tactile simulation and modeling continuously verify the data's physical consistency, ensuring recorded contact information obeys real-world physics and can be directly used as training samples.
Following the model's open-source release and real-world application, Xense Robotics says it will keep refining its tactile hardware and data infrastructure. Field testing by frontier teams has yielded crucial product feedback, clarifying which contact details matter and how perception should adapt to specific tasks. Going forward, Xense Robotics plans to fuel the steady evolution of robotic manipulation by supplying high-quality, real-world tactile data.
Gasgoo not only offers timely news and profound insight about China auto industry, but also help with business connection and expansion for suppliers and purchasers via multiple channels and methods. Buyer service: [email protected] Seller Service: [email protected]
All Rights Reserved. Do not reproduce, copy and use the editorial content without permission. Contact us: [email protected]
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported success rate
58.33%
View original evidence
al-tactile world model predicts the visual and tactile feedback after an action is executed, supporting reinforcement-learning post-training. Training runs in an interactive simulation environment before results are validated on real robots. Experiments show average real-robot success rates climbing from 58.33% to 75.00% for test-tube tasks, and from 80.00% to 86.67% for block-stacking tasks — gains of 16.67 and 6.67
Open source S5Reported success rate
75%
View original evidence
al-tactile world model predicts the visual and tactile feedback after an action is executed, supporting reinforcement-learning post-training. Training runs in an interactive simulation environment before results are validated on real robots. Experiments show average real-robot success rates climbing from 58.33% to 75.00% for test-tube tasks, and from 80.00% to 86.67% for block-stacking tasks — gains of 16.67 and 6.67
Open source S5Reported success rate
20%
View original evidence
tactile deformation, and force-related signals. Ablation studies reveal that adding this module doubled the success rate for cucumber peeling — from 20% to 40% — and lifted plate-wiping accuracy from 75% to 90%. For tasks demanding continuous surface contact, real-world interaction data gives action learning a direct grounding. During pre-deployment policy training, Anyverse Dynamics' in-house Phoenix-PhysCanvas visu
Open source S4Reported success rate
40%
View original evidence
tactile deformation, and force-related signals. Ablation studies reveal that adding this module doubled the success rate for cucumber peeling — from 20% to 40% — and lifted plate-wiping accuracy from 75% to 90%. For tasks demanding continuous surface contact, real-world interaction data gives action learning a direct grounding. During pre-deployment policy training, Anyverse Dynamics' in-house Phoenix-PhysCanvas visu
Open source S4Reported accuracy
75%
View original evidence
tactile deformation, and force-related signals. Ablation studies reveal that adding this module doubled the success rate for cucumber peeling — from 20% to 40% — and lifted plate-wiping accuracy from 75% to 90%. For tasks demanding continuous surface contact, real-world interaction data gives action learning a direct grounding. During pre-deployment policy training, Anyverse Dynamics' in-house Phoenix-PhysCanvas visu
Open source S4
- 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.
Gasgoo Munich- Anyverse Dynamics has released and open-sourced MWA ™ -Phoenix, a fully pre-trained, 7-billion-parameter latent-space world model foundation built for edge deployment. Public benchmarks show the model topping six simulation and real-robot leaderboards — including LIBERO-Plus, LIBERO-Pro, and the RoboDojo real-robot ranking. It hit a 94.63% overall success rate on LIBERO-Plus. Across all real-robot demo
Open source S2
al-tactile world model predicts the visual and tactile feedback after an action is executed, supporting reinforcement-learning post-training. Training runs in an interactive simulation environment before results are validated on real robots. Experiments show average real-robot success rates climbing from 58.33% to 75.00% for test-tube tasks, and from 80.00% to 86.67% for block-stacking tasks — gains of 16.67 and 6.67
Open source S5
tactile deformation, and force-related signals. Ablation studies reveal that adding this module doubled the success rate for cucumber peeling — from 20% to 40% — and lifted plate-wiping accuracy from 75% to 90%. For tasks demanding continuous surface contact, real-world interaction data gives action learning a direct grounding. During pre-deployment policy training, Anyverse Dynamics' in-house Phoenix-PhysCanvas visu
Open source S4
Source：Robotics — Robot learning news discovery · autonews.gasgoo.com