GIFT: Guided Intermediate Feature Training via Action-Oriented Structural Supervision for Robotic Manipulation

• URL: http://arxiv.org/abs/2609.04193v1
• Date: Thu, 03 Sep 2026 17:59:03 GMT
• Title: GIFT: Guided Intermediate Feature Training via Action-Oriented Structural Supervision for Robotic Manipulation
• Abstract summary: 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.
• Score: 60.95802802905149
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Vision-language pre-training and predictive world modeling provide robot policies with rich semantic and dynamic visual features, but their native action and visual-prediction objectives may omit critical physical and task structure while retaining control-irrelevant visual redundancy. We call this mismatch between visual richness and control utility the action-sufficiency gap. We investigate whether this gap can be bridged by guiding intermediate features to preserve three control-relevant structure in robotic manipulation: geometry governing motion feasibility, affordance encoding instruction-relevant entities, and goals grounding instructions in task-relevant regions. To this end, we present GIFT (Guided Intermediate Feature Training), an architecture-flexible framework for learning intermediate features that translates these structures into training-time constraints through geometry alignment, affordance prediction, and goal-region reconstruction. 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. Under zero-shot transfer to LIBERO-Plus, GIFT-VLA, GIFT-WAM-Fast, and GIFT-WAM-IDM outperform StarVLA-OFT, Fast-WAM, and Fast-WAM-IDM by 4.6, 12.6, and 5.2 points, reaching 79.6%, 72.6%, and 87.8%, respectively. 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. Together, these results establish learning functionally structured intermediate features as a reusable principle across model-specific action formulations, with especially large gains on articulated-object tasks and high-precision real-world manipulation under unseen visual and spatial perturbations. Project page: https://openphoenix-team.github.io/GIFT-pages.

Related papers

• AffordanceVLA: A Vision-Language-Action Model Empowering Action Generation through Affordance-Aware Understanding [55.535374902204296]
  AffordanceVLA is a unified framework that introduces structured affordance forecasting as a task-oriented intermediate representation.<n>We show that AffordanceVLA achieves strong performance across diverse manipulation scenarios.
  arXiv  Detail & Related papers  (2026-06-04T13:28:51Z)
• AttenA+: Rectifying Action Inequality in Robotic Foundation Models [38.61160855341111]
  We introduce AttenA+, an architecture-agnostic framework that prioritizes kinematically critical segments via velocity-driven action attention.<n>Our work suggests that mining the intrinsic structural priors of action sequences offers a highly efficient, physics-aware complement to standard scaling laws.
  arXiv  Detail & Related papers  (2026-05-13T13:55:37Z)
• RotVLA: Rotational Latent Action for Vision-Language-Action Model [54.22746299071677]
  We introduce RotVLA, a VLA framework built on a continuous rotational latent action representation.<n>Latent actions are modeled as elements of SO(n), providing continuity, compositionality, and structured geometry aligned with real-world action dynamics.<n>RotVLA consists of a VLM backbone and a flow-matching action head, pretrained on large-scale cross-embodiment robotic datasets and human videos.
  arXiv  Detail & Related papers  (2026-05-13T11:58:02Z)
• Beyond World-Frame Action Heads: Motion-Centric Action Frames for Vision-Language-Action Models [52.17964309142613]
  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.
  arXiv  Detail & Related papers  (2026-05-12T09:03:39Z)
• SG-VLA: Learning Spatially-Grounded Vision-Language-Action Models for Mobile Manipulation [43.77409401429281]
  Vision-Language-Action (VLA) models show promise for robotic control, yet performance in complex household environments remains sub-optimal.<n>Mobile manipulation requires reasoning about global scene layout, fine-grained geometry, and high-dimensional continuous actions.<n>We introduce a framework for learning spatially-grounded VLA models that strengthens perception and representation.
  arXiv  Detail & Related papers  (2026-03-24T03:44:25Z)
• Universal Pose Pretraining for Generalizable Vision-Language-Action Policies [83.39008378156647]
  Existing Vision-Language-Action (VLA) models often suffer from feature collapse and low training efficiency.<n>We propose Pose-VLA, a decoupled paradigm that separates VLA training into a pre-training phase for extracting universal 3D spatial priors.<n>Our framework follows a two-stage pre-training pipeline, establishing fundamental spatial grounding via poses followed by motion alignment.
  arXiv  Detail & Related papers  (2026-02-23T11:00:08Z)
• TC-IDM: Grounding Video Generation for Executable Zero-shot Robot Motion [38.88993836311246]
  Generative world models offer a promising alternative for general-purpose embodied AI.<n>We propose the Tool-Centric Inverse Dynamics Model (TC-IDM) to bridge the gap between visual planning and physical control.<n>In real-world evaluations, the world model with TC-IDM achieves an average success rate of 61.11 percent, with 77.7 percent on simple tasks and 38.46 percent on zero-shot deformable object tasks.
  arXiv  Detail & Related papers  (2026-01-26T10:06:56Z)
• Flex: End-to-End Text-Instructed Visual Navigation from Foundation Model Features [59.892436892964376]
  We investigate the minimal data requirements and architectural adaptations necessary to achieve robust closed-loop performance with vision-based control policies.<n>Our findings are synthesized in Flex (Fly lexically), a framework that uses pre-trained Vision Language Models (VLMs) as frozen patch-wise feature extractors.<n>We demonstrate the effectiveness of this approach on a quadrotor fly-to-target task, where agents trained via behavior cloning successfully generalize to real-world scenes.
  arXiv  Detail & Related papers  (2024-10-16T19:59:31Z)

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.