Geometric Parrots: Rule Absence vs. Rule Fragility in Spatial and Physical Reasoning for Vision-Language Models

arXiv:2610.16170 · cs.CV · Submitted 2026-10-12 · Read on arXiv

cs.CV

Submitted: 2026-10-12

Updated: 2026-10-12

Project page: https://bilaljawaid980.github.io/grip-benchmark

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Vision-language models — 1× in this paper · explained in 2 episodes

Sources

• PhysBench: Benchmarking and Enhancing Vision-Language Models for Physical World Understanding
• Why AI systems don't learn and what to do about it: Lessons on autonomous learning from cognitive science
• Why Language Models Hallucinate
• Discovering Language Model Behaviors with Model-Written Evaluations
• PENDULUM: A Benchmark for Assessing Sycophancy in Multimodal Large Language Models
• Vision language models are blind: Failing to translate detailed visual features into words
• Can Vision Language Models Learn Intuitive Physics from Interaction?
• Towards Understanding Sycophancy in Language Models
• The ARC of Progress towards AGI: A Living Survey of Abstraction and Reasoning
• Do Medical Vision Language Models Actually See? A Counterfactual Grounding Framework and Hard-Negative Contrastive Training for Visually-Reliant Medical VLMs

Related papers

• Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
• AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
• Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
• MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
• TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
• A Survey on Efficient Vision-Language-Action Models