V-CAGE: Context-Aware Generation and Verification for Scalable Long-Horizon Embodied Tasks

2026-01-21

AI Summary

This work addresses key challenges in long-horizon embodied intelligence learning—namely, physical implausibility, semantic inconsistency, and the difficulty of translating high-level instructions into executable actions in synthetic data. To this end, the authors propose V-CAGE, a closed-loop framework that ensures geometric consistency through context-aware scene instantiation, maps high-level goals to composable action primitives via hierarchical instruction decomposition, and incorporates a vision-language model (VLM)-driven verification loop for semantic validation. By integrating dynamic forbidden-region mapping with a rejection sampling mechanism, V-CAGE substantially enhances the physical and semantic fidelity of synthesized data. The resulting dataset significantly improves both the success rate and generalization capability of downstream embodied policies.

Technology Category

• Intelligent Robots: Embodied AI
• Computer Vision: Computational Photography, Image & Video Synthesis
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Learning long-horizon embodied behaviors from synthetic data remains challenging because generated scenes are often physically implausible, language-driven programs frequently"succeed"without satisfying task semantics, and high-level instructions require grounding into executable action sequences. To address these limitations, we introduce V-CAGE, a closed-loop framework for generating robust, semantically aligned manipulation datasets at scale. First, we propose a context-aware instantiation mechanism that enforces geometric consistency during scene synthesis. By dynamically maintaining a map of prohibited spatial areas as objects are placed, our system prevents interpenetration and ensures reachable, conflict-free configurations in cluttered environments. Second, to bridge the gap between abstract intent and low-level control, we employ a hierarchical instruction decomposition module. This decomposes high-level goals (e.g.,"get ready for work") into compositional action primitives, facilitating coherent long-horizon planning. Crucially, we enforce semantic correctness through a VLM-based verification loop. Acting as a visual critic, the VLM performs rigorous rejection sampling after each subtask, filtering out"silent failures"where code executes but fails to achieve the visual goal. Experiments demonstrate that V-CAGE yields datasets with superior physical and semantic fidelity, significantly boosting the success rate and generalization of downstream policies compared to non-verified baselines.

Problem

• long-horizon embodied tasks
• physically implausible scenes
• task semantics
• instruction grounding
• synthetic data generation

Innovation

• context-aware generation
• hierarchical instruction decomposition
• VLM-based verification
• long-horizon embodied tasks
• semantic fidelity