VeriSpace: Spatially Grounded Action Verification for Vision-Language-Action Models

2026-06-09

AI Summary

Existing vision-language-action (VLA) models rely solely on a single action prediction at test time, making them prone to irrecoverable minor yet critical errors that often lead to grasp failures or task deviations. This work proposes the first 3D-aware action verification mechanism, which constructs a scene representation fusing visual semantics and explicit geometry through a dual-path 3D-injected encoding scheme. Coupled with spatially anchored reasoning, the approach evaluates and selects among multiple candidate actions based on task relevance, geometric feasibility, and progress toward the goal. This framework substantially enhances discrimination of subtle action differences while remaining compatible with existing VLA policies. It consistently outperforms baseline models and current verification methods across both in-distribution and out-of-distribution scenarios on public benchmarks and real-world robotic tasks.

Abstract

Vision-language-action (VLA) models have shown strong promise for robotic manipulation, but their reliability at test time remains limited by one-shot action prediction, where even small action errors can cause grasp failure, collision, or incorrect task progression. A natural alternative is to equip VLA systems with test-time verification, allowing multiple candidate actions to be proposed and evaluated before execution. However, reliable action verification is challenging because it requires not only distinguishing subtle geometric differences between candidate actions, but also assessing whether an action makes meaningful progress toward the task goal. We present VeriSpace, a 3D-aware action verifier for test-time action selection in VLA systems. VeriSpace evaluates candidate actions through two key components: Dual-Path 3D-Injected Scene Encoding, which constructs a scene representation that jointly preserves visual semantics and explicit 3D geometry, and Spatially-Grounded Action Reasoning, which evaluates each action by reasoning over task-relevant spatial relations, geometric validity, and expected goal progress. Together, these components enable more reliable discrimination between subtle yet outcome-critical action candidates while remaining fully compatible with existing VLA policies. Experiments on public benchmarks and real-world robotic manipulation tasks show that VeriSpace consistently improves decision reliability over both underlying VLA policies and prior verification-based methods, yielding substantial gains in both in-distribution and out-of-distribution settings.

Problem

• action verification
• vision-language-action models
• robotic manipulation
• test-time decision
• spatial reasoning

Innovation

• 3D-aware verification
• spatially-grounded reasoning
• vision-language-action models
• action selection
• geometric validity