Beyond World-Frame Action Heads: Motion-Centric Action Frames for Vision-Language-Action Models

• URL: http://arxiv.org/abs/2605.11809v1
• Date: Tue, 12 May 2026 09:03:39 GMT
• Title: Beyond World-Frame Action Heads: Motion-Centric Action Frames for Vision-Language-Action Models
• Abstract summary: Vision-Language-Action (VLA) models have advanced rapidly with stronger backbones, broader pre-training, and larger demonstration datasets.<n>We propose textbfMCF-Proto, a lightweight action head that equips VLA policies with a Motion-Centric Action Frame (MCF) and a prototype-based action parameterization.<n>Our results suggest that adding lightweight geometric and compositional structure to the action head can materially improve how VLA policies organize and generalize robotic manipulation behavior.
• Score: 52.17964309142613
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Vision-Language-Action (VLA) models have advanced rapidly with stronger backbones, broader pre-training, and larger demonstration datasets, yet their action heads remain largely homogeneous: most directly predict action commands in a fixed world coordinate frame. We propose \textbf{MCF-Proto}, a lightweight action head that equips VLA policies with a Motion-Centric Action Frame (MCF) and a prototype-based action parameterization. At each step, the policy predicts a rotation $R_t \in SO(3)$, composes actions in the transformed local frame from a set of prototypes, and maps them back to the world frame for end-to-end training, using only standard demonstrations without auxiliary supervision. This simple design induces stable emergent structure. Without explicit directional labels, the learned local frames develop a stable geometric structure whose axes are strongly compatible with demonstrated end-effector motion. Meanwhile, actions in the learned representation become substantially more compact, with variation captured by fewer dominant directions and more regularly organized by shared prototypes. These structural properties translate into improved robustness, especially under geometric perturbations. Our results suggest that adding lightweight geometric and compositional structure to the action head can materially improve how VLA policies organize and generalize robotic manipulation behavior. An anonymized code repository is provided in the supplementary material.

Related papers

• GIFT: Guided Intermediate Feature Training via Action-Oriented Structural Supervision for Robotic Manipulation [60.95802802905149]
  We present GIFT (Guided Intermediate Feature Training), an architecture-flexible framework for learning intermediate features.<n>We instantiate GIFT in a Vision-Language-Action (VLA) policy, a direct-action World-Action Model (WAM), and an inverse-dynamics WAM while retaining each model's action formulation.<n>On RoboCasa, the three GIFT variants reach 61.4%, 83.6%, and 82.3%, outperforming their counterparts by 12.6, 9.0, and 8.4 points, respectively.
  arXiv  Detail & Related papers  (2026-09-03T17:59:03Z)
• D-CLOT: Double Closed Loop Optimal Transport for Unsupervised Action Segmentation [4.26168244766209]
  We build on the recently introduced CLOT to refine frame embeddings based on estimated segment embeddings.<n>We introduce a graph-constrained module that regularizes the OT-refined frame and segment representations.<n>Both variants improve segment-level quality over CLOT, with per-video gains of up to $+12.7$ F1 and $+10.2$ mIoU (YTI) and activity-level gains of up to $+8.9$ F1 (FS-Eval)
  arXiv  Detail & Related papers  (2026-08-06T10:59:28Z)
• Not All Actions Are Equal: Rethinking Conditioning for Dexterous World Model [15.646017114869169]
  Action conditioning remains underexplored in action-conditioned world models.<n>We propose DexAC, which treats action conditioning as a structured process rather than global compression.<n>We show that combining the semantic branch with DexAC significantly improves FID, FVD, and PCK, demonstrating gains in visual-temporal realism and action-following consistency.
  arXiv  Detail & Related papers  (2026-06-25T17:36:35Z)
• iMaC: Translating Actions into Motion and Contact Images for Embodied World Models [77.395425755122]
  Embodied world models have emerged as a pivotal paradigm for visual robotic decision-making and interactive environment simulation.<n>This paper proposesiMac (Image as Action Control), a novel unified control paradigm that treats raw visual images as native action representations for embodied world models.
  arXiv  Detail & Related papers  (2026-06-08T17:55:41Z)
• Lifting Embodied World Models for Planning and Control [59.09016913513998]
  We train a lightweight policy that maps high-level actions to sequences of low-level joint actions.<n>We instantiate this framework for a human-like embodiment, defining the high-level action space as a small set of 2D waypoints.<n>We show that the lifted world model substantially outperforms searching directly in low-level joint space.
  arXiv  Detail & Related papers  (2026-04-28T23:59:19Z)
• AnchorRefine: Synergy-Manipulation Based on Trajectory Anchor and Residual Refinement for Vision-Language-Action Models [60.04879435087352]
  Vision-language-action (VLA) policies generate actions within a single unified space.<n>We propose AnchorRefine, a hierarchical framework that factorizes VLA action modeling into trajectory anchor and residual refinement.<n> Experiments on LIBERO, CALVIN, and real-robot tasks demonstrate that AnchorRefine consistently improves both regression-based and diffusion-based VLA backbones.
  arXiv  Detail & Related papers  (2026-04-20T04:25:24Z)
• D-Prism: Differentiable Primitives for Structured Dynamic Modeling [46.22518390136366]
  D-Prism is the first framework to achieve high-fidelity structured dynamic modeling by extending differentiable primitives to the dynamic domain.<n>We introduce a deformation network to control primitive motion, ensuring it accurately matches the object's movement.<n>Experiments confirm that our method excels at structured dynamic modeling, providing both structured geometry and precise motion tracking.
  arXiv  Detail & Related papers  (2026-04-18T17:39:47Z)
• Beyond Dense Futures: World Models as Structured Planners for Robotic Manipulation [43.5447478385855]
  We propose StructVLA, which reformulates a generative world model into an explicit structured planner for reliable control.<n>We implement this approach through a two-stage training paradigm with a unified discrete token vocabulary.<n>In our experiments, StructVLA achieves strong average success rates of 75.0% on SimplerEnv-WidowX and 94.8% on LIBERO.
  arXiv  Detail & Related papers  (2026-03-13T01:33:48Z)
• Structural Action Transformer for 3D Dexterous Manipulation [80.07649565189035]
  Cross-embodiment skill transfer is a challenge for high-DoF robotic hands.<n>Existing methods, often relying on 2D observations and temporal-centric action representation, struggle to capture 3D spatial relations.<n>This paper proposes a new 3D dexterous manipulation policy that challenges this paradigm by introducing a structural-centric perspective.
  arXiv  Detail & Related papers  (2026-03-04T11:38:12Z)
• DynaPURLS: Dynamic Refinement of Part-aware Representations for Skeleton-based Zero-Shot Action Recognition [51.80782323686666]
  We introduce textbfDynaPURLS, a unified framework that establishes robust, multi-scale visual-semantic correspondences.<n>Our framework leverages a large language model to generate hierarchical textual descriptions that encompass both global movements and local body-part dynamics.<n>Experiments on three large-scale benchmark datasets, including NTU RGB+D 60/120 and PKU-MMD, demonstrate that DynaPURLS significantly outperforms prior art.
  arXiv  Detail & Related papers  (2025-12-12T10:39:10Z)
• Towards Efficient General Feature Prediction in Masked Skeleton Modeling [59.46799426434277]
  We propose a novel General Feature Prediction framework (GFP) for efficient mask skeleton modeling.<n>Our key innovation is replacing conventional low-level reconstruction with high-level feature prediction that spans from local motion patterns to global semantic representations.
  arXiv  Detail & Related papers  (2025-09-03T18:05:02Z)
• PMG: Progressive Motion Generation via Sparse Anchor Postures Curriculum Learning [5.247557449370603]
  ProMoGen is a novel framework that integrates trajectory guidance with sparse anchor motion control.<n>ProMoGen supports both dual and single control paradigms within a unified training process.<n>Our approach seamlessly integrates personalized motion with structured guidance, significantly outperforming state-of-the-art methods.
  arXiv  Detail & Related papers  (2025-04-23T13:51:42Z)
• POTLoc: Pseudo-Label Oriented Transformer for Point-Supervised Temporal Action Localization [26.506893363676678]
  This paper proposes POTLoc, a Pseudo-label Oriented Transformer for weakly-supervised Action localization. POTLoc is designed to identify and track continuous action structures via a self-training strategy. It outperforms the state-of-the-art point-supervised methods on THUMOS'14 and ActivityNet-v1.2 datasets.
  arXiv  Detail & Related papers  (2023-10-20T15:28:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.