aiqbench analysis
This arXiv preprint proposes a vision-language-action robot policy that learns from action-free human videos and predicts future visual features through a self-supervised EMA teacher, which is removed at inference. Its main claim is that explicit future prediction improves generalization and robustness, tested on LIBERO, LIBERO-Plus, SimplerEnv, and two real robots. These are the authors' own benchmark results, so readers should watch for independent replication and how the method compares with other world-model and latent-action approaches.
TS-WAM: Temporal Latent Actions and World-Action Coupling for Vision-Language Robot Control
arXiv:2610.21922v1 Announce Type: new Abstract: Many vision-language-action policies rely on action-labeled robot demonstrations and lack an explicit mechanism for predicting future visual states. These limitations hinder learning from action-free videos and can encourage overfitting to demonstration conditions, weakening generalization and robustness. We present TS-WAM, a lightweight Teacher-Student World-Action Model that addresses these challenges through self-supervised temporal learning and explicit future prediction. Its vlm-jepa framework learns from heterogeneous data, including robot demonstrations and action-free human videos, using an EMA target branch without requiring a separately trained external teacher. By coupling vlm-jepa with delta-lwm, TS-WAM learns to predict latent actions from the current observation and instruction and uses them to predict future visual features that guide action generation at inference. A unified end-to-end training architecture jointly optimizes temporal representation learning, visual dynamics, and robot control without a separate latent-action or world-model pretraining stage. The EMA target branch is removed at inference, keeping the deployed policy lightweight. Experiments on LIBERO, LIBERO-Plus, and SimplerEnv, together with real-world manipulation experiments on Franka Research 3 and RevoTron, demonstrate that TS-WAM achieves strong manipulation performance, generalization, and robustness.
Abstract:Many vision-language-action policies rely on action-labeled robot demonstrations and lack an explicit mechanism for predicting future visual states. These limitations hinder learning from action-free videos and can encourage overfitting to demonstration conditions, weakening generalization and robustness. We present TS-WAM, a lightweight Teacher-Student World-Action Model that addresses these challenges through self-supervised temporal learning and explicit future prediction. Its vlm-jepa framework learns from heterogeneous data, including robot demonstrations and action-free human videos, using an EMA target branch without requiring a separately trained external teacher. By coupling vlm-jepa with delta-lwm, TS-WAM learns to predict latent actions from the current observation and instruction and uses them to predict future visual features that guide action generation at inference. A unified end-to-end training architecture jointly optimizes temporal representation learning, visual dynamics, and robot control without a separate latent-action or world-model pretraining stage. The EMA target branch is removed at inference, keeping the deployed policy lightweight. Experiments on LIBERO, LIBERO-Plus, and SimplerEnv, together with real-world manipulation experiments on Franka Research 3 and RevoTron, demonstrate that TS-WAM achieves strong manipulation performance, generalization, and robustness.
Robotics (cs.RO)
arXiv:2610.21922 [cs.RO]
(or arXiv:2610.21922v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.21922 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 269 words
Related coverage
SLIP-VLA: Single-Step Latent Imagination for Policy Learning in Vision-Language-Action Models
arXiv · Sep 29, 2026
CoWorld-VLA: Thinking in a Multi-Expert World Model for Autonomous Driving
arXiv · Aug 27, 2026
Pretrained to Imagine, Fine-Tuned to Act: The Rise of World-Action Models
NVIDIA · Jun 15, 2026
LiMA: Bridging Long-term Imagination to Real-time Dexterous Manipulation via Asynchronous Diffusion
arXiv · Sep 24, 2026