aiqbench analysis
This arXiv preprint proposes a continuous latent-action policy that uses flow matching to produce executable robot actions without needing future video frames at inference. It is distilled from a larger latent-action teacher and reports gains over latent-action and vision-language-action baselines on hand-pose forecasting, manipulation, and real-world tests. For ML readers, the claims rest on a 1.1B-parameter model and benchmark results that still need independent replication, but the approach of learning transferable motion representations from unlabeled video is worth watching for robot learning.
UniLAP: A Unified Latent Action Policy for Cross-Embodiment Manipulation
arXiv:2610.16969v1 Announce Type: new Abstract: Latent Action Models (LAMs) can capture manipulation priors from action-free videos, but discrete latents usually impose a granularity bottleneck, limiting their ability to represent fine-grained action. Recent continuous alternatives mitigate this, but they are typically used as proxies for world modeling and require a further planning step to produce executable actions. We thus introduce UniLAP, a unified continuous latent-action policy that models the latent space via flow matching and bypasses reliance on future frames during inference. By distilling a LAM teacher with a flow-adaptive objective, UniLAP inherits the teacher's action representation while retaining the multimodal distribution of feasible behaviors. To decouple latent representations from task-irrelevant appearance, we extract a label-free motion target to isolate relevant kinematic cues from spurious visual dynamics. With the established embodiment-agnostic latent manifold, UniLAP can be easily adapted to novel platforms. Pretrained on multi-source videos at a compact 1.1B parameter scale, UniLAP outperforms state-of-the-art latent-action and Vision-Language-Action (VLA) baselines across hand pose forecasting, single-arm and humanoid manipulation, and real-world deployment.
Abstract:Latent Action Models (LAMs) can capture manipulation priors from action-free videos, but discrete latents usually impose a granularity bottleneck, limiting their ability to represent fine-grained action. Recent continuous alternatives mitigate this, but they are typically used as proxies for world modeling and require a further planning step to produce executable actions. We thus introduce UniLAP, a unified continuous latent-action policy that models the latent space via flow matching and bypasses reliance on future frames during inference. By distilling a LAM teacher with a flow-adaptive objective, UniLAP inherits the teacher's action representation while retaining the multimodal distribution of feasible behaviors. To decouple latent representations from task-irrelevant appearance, we extract a label-free motion target to isolate relevant kinematic cues from spurious visual dynamics. With the established embodiment-agnostic latent manifold, UniLAP can be easily adapted to novel platforms. Pretrained on multi-source videos at a compact 1.1B parameter scale, UniLAP outperforms state-of-the-art latent-action and Vision-Language-Action (VLA) baselines across hand pose forecasting, single-arm and humanoid manipulation, and real-world deployment.
Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2610.16969 [cs.RO]
(or arXiv:2610.16969v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.16969 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 · 263 words
Related coverage
HARP-VLA: Human-Robot Aligned Representation Learning for Vision-Language-Action Model
arXiv · Oct 12, 2026
Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
AWS · Sep 4, 2026
Meet Qwen-RobotSuite: Three Embodied AI Models for VLA Manipulation, Video World Modeling, and Navigation
MarkTechPost · Jun 16, 2026
Transferring Contact, Not Just Motion: Compliant Grasping Across Dexterous Hands
arXiv · Jun 16, 2026