aiqbench analysis
This research introduces TTL-SR, a test-time learning framework that enhances visual-language models' ability to perform quantitative spatial reasoning by leveraging geometric consistency constraints on unlabeled data. By adapting model parameters during inference without requiring expensive 3D supervision, the method significantly improves accuracy and robustness against distribution shifts in novel object configurations. The approach directly addresses critical limitations in current VLMs regarding their alignment with underlying geometric principles.
Geometry-Aware Test-Time Learning for Quantitative Spatial Reasoning
arXiv:2609.06004v1 Announce Type: cross Abstract: Quantitative spatial reasoning in visual-language models (VLMs) aims to infer spatial distances and directional relationships among objects in 3D space from a 2D image and a natural language query. Despite recent progress, VLM spatial reasoning remains brittle under distribution shifts, largely due to the high cost of 3D supervision. As a result, models often produce inconsistent or contradictory predictions when faced with novel object configurations or rephrased spatial queries, revealing a misalignment between learned representations and underlying geometry. To address this, we propose TTL-SR, a geometry-aware Test-Time Learning framework for quantitative Spatial Reasoning that leverages geometric consistency constraints and unlabeled test data to adapt models to target domains. Specifically, TTL-SR augments the input query with geometrically coupled auxiliary queries, filters unreliable predictions via adaptive geometric triggering to construct structured token-level pseudo-labels, and updates model parameters under a geometry-aware multi-objective loss using only test data. Experimental results demonstrate that TTL-SR significantly boosts spatial reasoning performance, yielding 6.47% and 9.41% accuracy gains for Qwen3-VL-4B-Instruct and SpatialRGPT-VILA-1.5-8B on Q-Spatial-ScanNet dataset, respectively.
Abstract:Quantitative spatial reasoning in visual-language models (VLMs) aims to infer spatial distances and directional relationships among objects in 3D space from a 2D image and a natural language query. Despite recent progress, VLM spatial reasoning remains brittle under distribution shifts, largely due to the high cost of 3D supervision. As a result, models often produce inconsistent or contradictory predictions when faced with novel object configurations or rephrased spatial queries, revealing a misalignment between learned representations and underlying geometry. To address this, we propose TTL-SR, a geometry-aware Test-Time Learning framework for quantitative Spatial Reasoning that leverages geometric consistency constraints and unlabeled test data to adapt models to target domains. Specifically, TTL-SR augments the input query with geometrically coupled auxiliary queries, filters unreliable predictions via adaptive geometric triggering to construct structured token-level pseudo-labels, and updates model parameters under a geometry-aware multi-objective loss using only test data. Experimental results demonstrate that TTL-SR significantly boosts spatial reasoning performance, yielding 6.47% and 9.41% accuracy gains for Qwen3-VL-4B-Instruct and SpatialRGPT-VILA-1.5-8B on Q-Spatial-ScanNet dataset, respectively.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2609.06004 [cs.CV]
(or arXiv:2609.06004v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.06004 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 268 words
Related coverage
PIRAMID: Progress and Plans
LessWrong · Jul 27, 2026
Game Design Creativity evals
LessWrong · Oct 12, 2026
VeriSpace: Spatially Grounded Action Verification for Vision-Language-Action Models
arXiv · Jun 10, 2026
Test-Time Spatial Reasoning for Robot Manipulation Using Generative Real-to-Sim
arXiv · Sep 29, 2026