Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models

• URL: http://arxiv.org/abs/2609.12641v1
• Date: Fri, 11 Sep 2026 09:42:44 GMT
• Title: Breaking the Vision-Action Shortcut: Latent Interface Training for Generalizable Robotics Foundation Models
• Abstract summary: We propose a framework-agnostic two-stage strategy for training robot foundation models.<n> Stage 1 establishes a spatial-goal-conditioned action prior to images, then constrains visual conditioning through a pose-supervised latent interface.<n>Stage 2 introduces a latent interface that aggregates visual and semantic representations and serves as the pretrained action expert's only visual conditioning pathway.
• Score: 3.215979391291133
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Robot foundation models achieve strong in-distribution performance but often degrade under visual distribution shifts. When learning to generate actions from pretrained visual representations, models may exploit task-irrelevant visual cues that correlate with demonstrated actions within the training distribution. Such vision-action shortcuts can undermine generalization when these correlations change under distribution shifts. Mitigating these shortcuts requires constraining how visual information is used for action generation while preserving task-relevant spatial information. We propose Latent Interface Training (LIT), a framework-agnostic two-stage strategy that first establishes a spatial-goal-conditioned action prior without images, then constrains visual conditioning through a pose-supervised latent interface. Stage 1 trains the action expert to generate action chunks conditioned on language, robot state, and each demonstrated chunk's terminal SE(3) end-effector pose, learning goal-directed action generation independently of visual cues. Stage 2 introduces a latent interface that aggregates visual and semantic representations and serves as the pretrained action expert's only visual conditioning pathway. The interface is supervised to reconstruct the terminal pose previously used to condition Stage 1, encouraging it to retain the goal-relevant spatial information needed for action generation. Across four vision-language-action and world-action architectures (Pi0.5, MolmoAct2, FAST-WAM, and ImageWAM), LIT improves overall LIBERO-Plus success by 3.87-10.70 percentage points while preserving or improving average LIBERO success. Real-world evaluations show 13.30-16.70 percentage-point gains in success aggregated across three tasks under unseen camera configurations, lighting variations, and distractors.

Related papers

• LAWM-3D: Learning 3D-Aware Latent Actions from Human Videos for Generalizable Robot World Models [64.11117186468533]
  World models enable agents to perform forward rollout and planning without real-world interaction.<n>Most existing LAMs rely on single-view inputs and operate primarily in 2D pixel space.<n>We propose LAWM-3D, which introduces three tightly coupled key designs for learning 3D-aware latent actions.
  arXiv  Detail & Related papers  (2026-08-06T07:47:01Z)
• 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)
• Action with Visual Primitives [36.230909065494345]
  Vision-Language-Action (VLA) models have emerged as a promising paradigm for generalist robotic manipulation.<n>We introduce AVP (Action with Visual Primitives), an end-to-end architecture that implements this visual-primitive-centric interface.<n>We show that AVP improves the success rate by 27.61% over pi_0.5 and outperforms other recent methods.
  arXiv  Detail & Related papers  (2026-05-21T08:52:47Z)
• DiffusionAnything: End-to-End In-context Diffusion Learning for Unified Navigation and Pre-Grasp Motion [2.061143628317803]
  Recent vision-language-action (VLA) models infer actions directly from visual input but demand massive computational resources.<n>We present a unified image-space diffusion policy handling both meter-scale navigation and centimeter-scale manipulation.<n>The model achieves robust zero-shot generalization to novel scenes while remaining suitable for onboard deployment.
  arXiv  Detail & Related papers  (2026-03-27T11:40:13Z)
• 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)
• Agentic Jigsaw Interaction Learning for Enhancing Visual Perception and Reasoning in Vision-Language Models [63.69856480318313]
  AGILE formulates jigsaw solving as an interactive process, enabling the model to progressively engage with the environment.<n>We show that AGILE substantially boosts performance on jigsaw tasks of varying complexity.<n>We also demonstrate strong generalization across 9 general vision tasks, achieving an average improvement of 3.1%.
  arXiv  Detail & Related papers  (2025-10-01T17:58:05Z)
• 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)
• Learning View-Disentangled Human Pose Representation by Contrastive Cross-View Mutual Information Maximization [33.36330493757669]
  We introduce a novel representation learning method to disentangle pose-dependent as well as view-dependent factors from 2D human poses. The method trains a network using cross-view mutual information (CV-MIM) which maximizes mutual information of the same pose performed from different viewpoints. CV-MIM outperforms other competing methods by a large margin in the single-shot cross-view setting.
  arXiv  Detail & Related papers  (2020-12-02T18:55:35Z)
• Learning Long-term Visual Dynamics with Region Proposal Interaction Networks [75.06423516419862]
  We build object representations that can capture inter-object and object-environment interactions over a long-range. Thanks to the simple yet effective object representation, our approach outperforms prior methods by a significant margin.
  arXiv  Detail & Related papers  (2020-08-05T17:48:00Z)

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.