aiqbench analysis
This research introduces Context-Aware Motion Priors (CMP), a framework that dynamically adapts general motion priors to specific task contexts for humanoid control without requiring manual skill labels or separate discovery stages. By reweighting reference supervision based on learned context-motion compatibility, the method significantly improves sample efficiency and performance across diverse tasks while maintaining robustness against imbalanced datasets. The approach represents a meaningful advancement in leveraging prior knowledge more effectively within reinforcement learning policies for complex robotic manipulation.
Learning Context-Aware Motion Priors for Humanoid Control
arXiv:2608.03234v1 Announce Type: new Abstract: Motion priors provide powerful guidance for learning naturalistic humanoid behaviors. However, existing methods typically learn a general, task-agnostic prior from the entire reference dataset and apply it uniformly throughout policy training. As a result, the prior cannot distinguish which reference motions are relevant to the current task context, potentially providing irrelevant or conflicting guidance. We present Context-Aware Motion Priors (CMP), a framework that adapts a general motion prior to the current task context without manual skill labels, dataset partitioning, or a separate skill discovery stage. Specifically, CMP learns context-motion compatibility using high-advantage policy rollouts, while a demonstration-based objective keeps the learned relevance grounded in the reference distribution. The resulting relevance scores reweight reference supervision for training a lightweight context-conditioned adapter. To evaluate the effectiveness and generality of CMP, we instantiate it with both Adversarial Motion Priors and Score-Matching Motion Priors. Across five humanoid control tasks, CMP consistently improves task performance and sample efficiency, learns meaningful context-motion alignment, and remains robust to imbalanced reference distributions. These results show that adapting motion priors to task contexts provides more relevant guidance for humanoid policy learning.
Abstract:Motion priors provide powerful guidance for learning naturalistic humanoid behaviors. However, existing methods typically learn a general, task-agnostic prior from the entire reference dataset and apply it uniformly throughout policy training. As a result, the prior cannot distinguish which reference motions are relevant to the current task context, potentially providing irrelevant or conflicting guidance. We present Context-Aware Motion Priors (CMP), a framework that adapts a general motion prior to the current task context without manual skill labels, dataset partitioning, or a separate skill discovery stage. Specifically, CMP learns context-motion compatibility using high-advantage policy rollouts, while a demonstration-based objective keeps the learned relevance grounded in the reference distribution. The resulting relevance scores reweight reference supervision for training a lightweight context-conditioned adapter. To evaluate the effectiveness and generality of CMP, we instantiate it with both Adversarial Motion Priors and Score-Matching Motion Priors. Across five humanoid control tasks, CMP consistently improves task performance and sample efficiency, learns meaningful context-motion alignment, and remains robust to imbalanced reference distributions. These results show that adapting motion priors to task contexts provides more relevant guidance for humanoid policy learning.
16 pages, including appendices. Code will be released publicly
Robotics (cs.RO)
arXiv:2608.03234 [cs.RO]
(or arXiv:2608.03234v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2608.03234 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 · 284 words
Related coverage
Unifying Physics-Based Humanoid Interaction with a Context-Conditioned Interaction Prior
arXiv · Sep 9, 2026
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · Oct 13, 2026
Perceptive Behavior Foundation Model: Adapting Human Motion Priors to Robot-Centric Terrain
arXiv · Jun 16, 2026
Perceptive Behavior Foundation Model: Adapting Human Motion Priors to Robot-Centric Terrain
arXiv · Jun 9, 2026