# Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures

arXiv: https://arxiv.org/abs/2609.00663 · HTML: https://mlsanity.com/paper/2609.00663/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 25 of its 45 references that are in the map.
- First posted on arXiv 1 Sep 2026.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)

## Builds on (most cited first, 25 of 25)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [MVTN: Multi-View Transformation Network for 3D Shape Recognition](https://mlsanity.com/paper/2011.13244/) (2020)
- [Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens](https://mlsanity.com/paper/2511.19418/) (2025)
- [Identifiability Results for Multimodal Contrastive Learning](https://mlsanity.com/paper/2303.09166/) (2023)
- [ZeroBench: An Impossible Visual Benchmark for Contemporary Large Multimodal Models](https://mlsanity.com/paper/2502.09696/) (2025)
- [VisOnlyQA: Large Vision Language Models Still Struggle with Visual Perception of Geometric Information](https://mlsanity.com/paper/2412.00947/) (2024)
- [More Thought, Less Accuracy? On the Dual Nature of Reasoning in Vision-Language Models](https://mlsanity.com/paper/2509.25848/) (2025)
- [Seeing from Another Perspective: Evaluating Multi-View Understanding in MLLMs](https://mlsanity.com/paper/2504.15280/) (2025)
- [Visually Descriptive Language Model for Vector Graphics Reasoning](https://mlsanity.com/paper/2404.06479/) (2024)
- [Native Visual Understanding: Resolving Resolution Dilemmas in Vision-Language Models](https://mlsanity.com/paper/2506.12776/) (2025)
- [Res-Bench: Benchmarking the Robustness of Multimodal Large Language Models to Dynamic Resolution Input](https://mlsanity.com/paper/2510.16926/) (2025)
- [CRYSPNet: Crystal Structure Predictions via Neural Network](https://mlsanity.com/paper/2003.14328/) (2020)
- [What MLLMs Learn about When they Learn about Multimodal Reasoning](https://mlsanity.com/paper/2510.01719/) (2025)
- [Chain-of-Thought Degrades Visual Spatial Reasoning Capabilities of Multimodal LLMs](https://mlsanity.com/paper/2604.16060/) (2026)
- [From Seeing to Thinking: Decoupling Perception and Reasoning Improves Post-Training of Vision-Language Models](https://mlsanity.com/paper/2605.20177/) (2026)
- [Your Reasoning Benchmark May Not Test Reasoning: Revealing Perception Bottleneck in Abstract Reasoning Benchmarks](https://mlsanity.com/paper/2512.21329/) (2025)
- [Stress-Testing Multimodal Foundation Models for Crystallographic Reasoning](https://mlsanity.com/paper/2506.13051/) (2025)
- [Grounding the Score: Explicit Visual Premise Verification for Reliable Vision-Language Process Reward Models](https://mlsanity.com/paper/2603.16253/) (2026)
- [Unsolvability Ceiling in Multi-LLM Routing: An Empirical Study of Evaluation Artifacts](https://mlsanity.com/paper/2605.07395/) (2026)
- [CrystalXRD-Bench: Benchmarking Vision-Language Models for XRD Peak Indexing Across Diverse Crystalline Materials](https://mlsanity.com/paper/2605.29446/) (2026)
- [Visual Access Boundaries in Vision-Language Model Reasoning](https://mlsanity.com/paper/2607.12815/) (2026)
- [Designing a Dataset for Convolutional Neural Networks to Predict Space Groups Consistent with Extinction Laws](https://mlsanity.com/paper/2411.00803/) (2024)
- [M3-Verse: A "Spot the Difference" Challenge for Large Multimodal Models](https://mlsanity.com/paper/2512.18735/) (2025)

## Abstract (arXiv)

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.