2026 · arXiv · Preprint
Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures
Can Polat , Mustafa Kurban , Erchin Serpedin , Hasan KurbanResearch highlights
A render ceiling for separating perception and reasoning
Determine what a rendered crystal image supports before attributing a model's errors to reasoning.
Why we investigated it
A second learned model can introduce its own errors when used to separate perception from reasoning.
What the work contributes
The render ceiling uses known cameras and cross-view correspondence to recover an image-supported reference without a learned judge.
What the evidence shows
Across fourteen vision-language models, exact geometry improves performance but closes less than half the gap for thirteen. A supervised vision model exceeds every evaluated VLM.
Build on this work
Use the geometric reference and camera-design principles when building benchmarks with an invertible rendering process.
- Vision-language evaluation
- Crystal structures
- Perception
- Scientific reasoning
Abstract
Multimodal evaluations cannot say whether a vision-language model misread an image or misreasoned about it, because every existing method for separating the two places a second model in the loop. We introduce the render ceiling, a model-free reference for benchmarks built by rendering known objects: inverting the frozen cameras and re-solving cross-view correspondence recovers exactly the answer the images support. We prove the ceiling fails only through an enumerable set of projection coincidences and certify that set empty on 2,160 rendered crystal structures, so every point of a model's deficit belongs to the model. Across fourteen vision-language models, supplying exact geometry as text lifts every model yet closes under half the gap for thirteen, while a supervised vision model with no language component reads the same images at 0.8952, above every vision-language model. The instrument exposes extraction-stage fabrication that downstream accuracy would misattribute to reasoning, yields camera-placement rules for benchmark builders, and transfers to any benchmark with an invertible forward rendering.
Cite
@misc{polat2026separating,
title = {Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures},
author = {Can Polat and Mustafa Kurban and Erchin Serpedin and Hasan Kurban},
year = 2026,
journal = {arXiv},
doi = {10.48550/arxiv.2609.00663},
url = {https://arxiv.org/pdf/2609.00663},
}