aiqbench analysis
This paper introduces Proprioceptive Action Models (PAM), a novel approach for generative robot policies that jointly predicts compact, timing-free geometric sketches and dense executable action chunks within a single transformer. By parameterizing paths by arc length rather than time, PAM effectively captures long-horizon intent without the computational overhead of language plans or video forecasts, significantly improving success rates on both simulated and real-world bimanual tasks. The method demonstrates how advanced generative modeling can bridge the gap between high-level geometric planning and low-level motor execution in autonomous systems.
Proprioceptive Sketches as Long-Horizon Intent for Generative Action Policies
arXiv:2610.02759v1 Announce Type: new Abstract: Generative robot policies predict short action chunks but lack explicit long-horizon intent. Recent methods expose longer-horizon structure through language plans, subgoal images, or video forecasts, which are costly to generate and still need to be translated into robot motion. Predicting future robot motions avoids this translation, but a dense, time-indexed trajectory requires numerous parameters to cover the full remaining task, and over a short horizon it largely repeats the action chunk and adds little guidance for action generation. We propose Proprioceptive Action Models (PAM), which jointly generate a compact, timing-free sketch of the robot's remaining joint-space path and a dense executable action chunk within a single transformer denoiser. The sketch parameterizes the path by arc length rather than time, capturing geometric intent invariant to execution timing. Block-causal attention and a staggered denoising schedule maintain directed sketch-to-action dependence, ensuring the action tokens condition on a progressively cleaner sketch throughout sampling. In simulation, PAM improves over its action-only counterparts on Push-T and LIBERO-Long; on four real-world bimanual tasks, it raises success from 47.5% to 75.0%. Project page: https://nicehiro.github.io/pam_dp/
Abstract:Generative robot policies predict short action chunks but lack explicit long-horizon intent. Recent methods expose longer-horizon structure through language plans, subgoal images, or video forecasts, which are costly to generate and still need to be translated into robot motion. Predicting future robot motions avoids this translation, but a dense, time-indexed trajectory requires numerous parameters to cover the full remaining task, and over a short horizon it largely repeats the action chunk and adds little guidance for action generation. We propose Proprioceptive Action Models (PAM), which jointly generate a compact, timing-free sketch of the robot's remaining joint-space path and a dense executable action chunk within a single transformer denoiser. The sketch parameterizes the path by arc length rather than time, capturing geometric intent invariant to execution timing. Block-causal attention and a staggered denoising schedule maintain directed sketch-to-action dependence, ensuring the action tokens condition on a progressively cleaner sketch throughout sampling. In simulation, PAM improves over its action-only counterparts on Push-T and LIBERO-Long; on four real-world bimanual tasks, it raises success from 47.5% to 75.0%. Project page: this https URL
19 pages. Project page: this https URL
Robotics (cs.RO)
arXiv:2610.02759 [cs.RO]
(or arXiv:2610.02759v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.02759 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 · 278 words
Related coverage
Building a Foundation Stack for General-Purpose Robots
IEEE Spectrum · Jul 13, 2026
[AINews] OpenAI reports Navier-Stokes singularity find in 88 hours using Astra-next, roughly 10,000 agents and 130B tokens (>$40M), a contender for second ever Millennium Prize awarded
Latent Space · Sep 9, 2026
CoRE-VLA: Towards Scalable and Robust Vision-Language-Action Modeling via Conditional Routing of Experts
arXiv · Oct 12, 2026
How Telcos Build Autonomous Networks with Agentic AI
NVIDIA · Jun 23, 2026