aiqbench analysis
GeoAlign introduces a state-guided spatial alignment architecture that enhances Vision-Language-Action models by integrating RGB-D geometry features directly into policy learning. This approach significantly improves performance on geometry-critical real-world manipulation tasks, addressing the current limitation where VLA models prioritize semantic grounding over executable spatial accuracy. The method's efficacy is validated through strong results on standard benchmarks and proprietary datasets, confirming the necessity of proprioceptive-state-guided querying for robust robotic control.
GeoAlign: Beyond Semantics with State-Guided Spatial Alignment in VLA Models
arXiv:2606.03240v1 Announce Type: new Abstract: Current Vision--Language--Action (VLA) models often optimize for semantic grounding, whereas executable manipulation requires geometry-aware spatial alignment and dynamic affordance selection. We introduce GeoAlign, a state-guided spatial alignment architecture for VLA policy learning. GeoAlign post-trains an RGB geometry branch with robot-domain RGB-D supervision, yielding RGB-derived Geometry-Enhanced Post-Trained (GEP) features for policy rollout. The robot's proprioceptive state queries the GEP feature grid, producing compact, phase-dependent geometry tokens for action prediction. GeoAlign achieves 99.0% on LIBERO, 85.3% across three SimplerEnv-Fractal tasks, and 78.8% on eight geometry-critical real-world ALOHA tasks, with ablations confirming the value of geometry post-training and proprioceptive-state-guided querying.
Abstract:Current Vision--Language--Action (VLA) models often optimize for semantic grounding, whereas executable manipulation requires geometry-aware spatial alignment and dynamic affordance selection. We introduce GeoAlign, a state-guided spatial alignment architecture for VLA policy learning. GeoAlign post-trains an RGB geometry branch with robot-domain RGB-D supervision, yielding RGB-derived Geometry-Enhanced Post-Trained (GEP) features for policy rollout. The robot's proprioceptive state queries the GEP feature grid, producing compact, phase-dependent geometry tokens for action prediction. GeoAlign achieves 99.0% on LIBERO, 85.3% across three SimplerEnv-Fractal tasks, and 78.8% on eight geometry-critical real-world ALOHA tasks, with ablations confirming the value of geometry post-training and proprioceptive-state-guided querying.
Robotics (cs.RO)
arXiv:2606.03240 [cs.RO]
(or arXiv:2606.03240v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2606.03240 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 114 words
Related coverage
GeoAlign: Beyond Semantics with State-Guided Spatial Alignment in VLA Models
arXiv · Oct 7, 2026
Beyond Isolated Entities: Relation-Aware Multi-Entity Modeling for Unsupervised Video Anomaly Detection
arXiv · Sep 29, 2026
TacWAM: Anchor-Guided World Action Model with Mechanics-Aware Tactile Prediction
arXiv · Jul 31, 2026
Article: The AI Productivity Paradox in Test Automation: Moving beyond Structural Validation to Perception and Intent
InfoQ · Jun 1, 2026