# Geometric Perception on GePBench 1.0 (Hard)

Source: [Geometric Perception on GePBench 1.0 (Hard)](<https://www.sota2.com/research/sota/geometric-perception-on-gepbench-1-0-hard>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-16T09:10:53.903Z

Evaluation of geometric perception capabilities on the hard subset of the GePBench 1.0 dataset, assessing sub-tasks including existence, counting, size, location, reference, and relationship.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Geometric Perception |
| Dataset (including any stated split) | GePBench 1.0 (Hard) |
| Metric | Existence |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 34 of 34 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Existence](<https://www.sota2.com/research/sota/geometric-perception-on-gepbench-1-0-hard.md?metric=Existence>): Higher is better; unit: PERCENT.

- [Counting](<https://www.sota2.com/research/sota/geometric-perception-on-gepbench-1-0-hard.md?metric=Counting>): Higher is better; unit: PERCENT.

- [Size](<https://www.sota2.com/research/sota/geometric-perception-on-gepbench-1-0-hard.md?metric=Size>): Higher is better; unit: PERCENT.

- [Location](<https://www.sota2.com/research/sota/geometric-perception-on-gepbench-1-0-hard.md?metric=Location>): Higher is better; unit: PERCENT.

- [Reference](<https://www.sota2.com/research/sota/geometric-perception-on-gepbench-1-0-hard.md?metric=Reference>): Higher is better; unit: PERCENT.

- [Relationship](<https://www.sota2.com/research/sota/geometric-perception-on-gepbench-1-0-hard.md?metric=Relationship>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Existence | Result date | Report paper |
| --- | --- | --- | --- |
| Human | 99.3 | Not recorded | 68fa77071e209131cea709bb |
| GPT-5.2 | 81 | Not recorded | 68fa77071e209131cea709bb |
| Qwen3-VL | 79.5 | Not recorded | 68fa77071e209131cea709bb |
| Qwen3.5 | 74.9 | Not recorded | 68fa77071e209131cea709bb |
| InternVL3 | 73.8 | Not recorded | 68fa77071e209131cea709bb |
| GPT-4o | 73.3 | Not recorded | 68fa77071e209131cea709bb |
| InternVL3 | 73.3 | Not recorded | 68fa77071e209131cea709bb |
| Qwen2.5-VL | 71.3 | Not recorded | 68fa77071e209131cea709bb |
| InternVL2.5 | 71.3 | Not recorded | 68fa77071e209131cea709bb |
| Gemini-2.5-pro | 70.8 | Not recorded | 68fa77071e209131cea709bb |
| Qwen3.5 | 70.8 | Not recorded | 68fa77071e209131cea709bb |
| Qwen2-VL | 69.2 | Not recorded | 68fa77071e209131cea709bb |
| Qwen3-VL | 69.2 | Not recorded | 68fa77071e209131cea709bb |
| LLaVA-OneVision | 68.7 | Not recorded | 68fa77071e209131cea709bb |
| QVQ-preview | 67.2 | Not recorded | 68fa77071e209131cea709bb |
| Qwen2.5-VL | 66.7 | Not recorded | 68fa77071e209131cea709bb |
| MiniCPM-V-2.6 | 65.1 | Not recorded | 68fa77071e209131cea709bb |
| Qwen2-VL | 65.1 | Not recorded | 68fa77071e209131cea709bb |
| InternVL2.5 | 62.6 | Not recorded | 68fa77071e209131cea709bb |
| LLaMA-3.2-Vision | 60 | Not recorded | 68fa77071e209131cea709bb |
| LLaVA-OneVision | 60 | Not recorded | 68fa77071e209131cea709bb |
| mPLUG-Owl3 | 57.4 | Not recorded | 68fa77071e209131cea709bb |
| Mantis-Idefics2 | 55.9 | Not recorded | 68fa77071e209131cea709bb |
| GLM-4V | 53.3 | Not recorded | 68fa77071e209131cea709bb |
| Math-LLaVA | 46.2 | Not recorded | 68fa77071e209131cea709bb |
| Math-PUMA | 45.1 | Not recorded | 68fa77071e209131cea709bb |
| InstructBLIP | 42.1 | Not recorded | 68fa77071e209131cea709bb |
| LLaVA-1.5 | 42.1 | Not recorded | 68fa77071e209131cea709bb |
| BLIP2 | 41 | Not recorded | 68fa77071e209131cea709bb |
| LLaVA-1.5 | 32.8 | Not recorded | 68fa77071e209131cea709bb |
| G-LLaVA | 29.2 | Not recorded | 68fa77071e209131cea709bb |
| MiniGPTv2 | 28.2 | Not recorded | 68fa77071e209131cea709bb |
| Random guessing | 25 | Not recorded | 68fa77071e209131cea709bb |
| G-LLaVA | 22.1 | Not recorded | 68fa77071e209131cea709bb |

### Human

Existence: 99.3. Result ID: 6a31130bfbc373592c2d9ec3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"-","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### GPT-5.2

Existence: 81. Result ID: 6a31130bfbc373592c2d9ec5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"-","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Qwen3-VL

Existence: 79.5. Result ID: 6a31130bfbc373592c2d9ed8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6927db54d6b0dd3342930f66.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"235B-A22B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Qwen3.5

Existence: 74.9. Result ID: 6a31130bfbc373592c2d9eda.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"397B-A17B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### InternVL3

Existence: 73.8. Result ID: 6a31130bfbc373592c2d9edd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78241e209131cea76fa5.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"8B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### GPT-4o

Existence: 73.3. Result ID: 6a31130bfbc373592c2d9ec4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa727a1e209131cea56a1e.md>)

[Implementation](<https://github.com/openai/evals>)

Recorded evaluation configuration: {"Model Size (Parameters)":"-","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### InternVL3

Existence: 73.3. Result ID: 6a31130bfbc373592c2d9ede.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78241e209131cea76fa5.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"78B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Qwen2.5-VL

Existence: 71.3. Result ID: 6a31130bfbc373592c2d9ed6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"72B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### InternVL2.5

Existence: 71.3. Result ID: 6a31130bfbc373592c2d9edc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bab1e209131cea89c7f.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"Model Size (Parameters)":"78B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Gemini-2.5-pro

Existence: 70.8. Result ID: 6a31130bfbc373592c2d9ec6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78731e209131cea78622.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"-","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Qwen3.5

Existence: 70.8. Result ID: 6a31130bfbc373592c2d9ed9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"9B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Qwen2-VL

Existence: 69.2. Result ID: 6a31130bfbc373592c2d9ed4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75151e209131cea65976.md>)

[Implementation](<https://github.com/QwenLM/Qwen2-VL>)

Recorded evaluation configuration: {"Model Size (Parameters)":"72B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Qwen3-VL

Existence: 69.2. Result ID: 6a31130bfbc373592c2d9ed7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6927db54d6b0dd3342930f66.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"8B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### LLaVA-OneVision

Existence: 68.7. Result ID: 6a31130bfbc373592c2d9ed2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756f1e209131cea67e58.md>)

[Implementation](<https://llava-vl.github.io/blog/llava-onevision>)

Recorded evaluation configuration: {"Model Size (Parameters)":"72B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### QVQ-preview

Existence: 67.2. Result ID: 6a31130bfbc373592c2d9ee3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"72B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Qwen2.5-VL

Existence: 66.7. Result ID: 6a31130bfbc373592c2d9ed5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"7B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### MiniCPM-V-2.6

Existence: 65.1. Result ID: 6a31130bfbc373592c2d9ecd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74681e209131cea6183d.md>)

[Implementation](<https://github.com/OpenBMB/MiniCPM-V>)

Recorded evaluation configuration: {"Model Size (Parameters)":"8B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Qwen2-VL

Existence: 65.1. Result ID: 6a31130bfbc373592c2d9ed3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75151e209131cea65976.md>)

[Implementation](<https://github.com/QwenLM/Qwen2-VL>)

Recorded evaluation configuration: {"Model Size (Parameters)":"7B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### InternVL2.5

Existence: 62.6. Result ID: 6a31130bfbc373592c2d9edb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bab1e209131cea89c7f.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"Model Size (Parameters)":"8B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### LLaMA-3.2-Vision

Existence: 60. Result ID: 6a31130bfbc373592c2d9ed0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"90B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### LLaVA-OneVision

Existence: 60. Result ID: 6a31130bfbc373592c2d9ed1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756f1e209131cea67e58.md>)

[Implementation](<https://llava-vl.github.io/blog/llava-onevision>)

Recorded evaluation configuration: {"Model Size (Parameters)":"7B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### mPLUG-Owl3

Existence: 57.4. Result ID: 6a31130bfbc373592c2d9ecc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa746c1e209131cea61a16.md>)

[Implementation](<https://github.com/X-PLUG/mPLUG-Owl>)

Recorded evaluation configuration: {"Model Size (Parameters)":"7B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Mantis-Idefics2

Existence: 55.9. Result ID: 6a31130bfbc373592c2d9ecf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75b11e209131cea6954e.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"8B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### GLM-4V

Existence: 53.3. Result ID: 6a31130bfbc373592c2d9ece.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74521e209131cea610b6.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"9B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Math-LLaVA

Existence: 46.2. Result ID: 6a31130bfbc373592c2d9ee1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75201e209131cea65fb5.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"13B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Math-PUMA

Existence: 45.1. Result ID: 6a31130bfbc373592c2d9ee2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74f81e209131cea64fc2.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"7B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### InstructBLIP

Existence: 42.1. Result ID: 6a31130bfbc373592c2d9ec8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fb71e209131cea47021.md>)

[Implementation](<https://github.com/salesforce/LAVIS/tree/main/projects/instructblip>)

Recorded evaluation configuration: {"Model Size (Parameters)":"3B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### LLaVA-1.5

Existence: 42.1. Result ID: 6a31130bfbc373592c2d9ecb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732e1e209131cea5af59.md>)

[Implementation](<https://llava-vl.github.io/>)

Recorded evaluation configuration: {"Model Size (Parameters)":"13B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### BLIP2

Existence: 41. Result ID: 6a31130bfbc373592c2d9ec7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fb71e209131cea46fe6.md>)

[Implementation](<https://github.com/salesforce/LAVIS/tree/main/projects/blip2>)

Recorded evaluation configuration: {"Model Size (Parameters)":"3B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### LLaVA-1.5

Existence: 32.8. Result ID: 6a31130bfbc373592c2d9eca.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa732e1e209131cea5af59.md>)

[Implementation](<https://llava-vl.github.io/>)

Recorded evaluation configuration: {"Model Size (Parameters)":"7B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### G-LLaVA

Existence: 29.2. Result ID: 6a31130bfbc373592c2d9ee0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ac11e209131cea8503e.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"13B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### MiniGPTv2

Existence: 28.2. Result ID: 6a31130bfbc373592c2d9ec9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71251e209131cea4f7e5.md>)

[Implementation](<https://minigpt-v2.github.io/>)

Recorded evaluation configuration: {"Model Size (Parameters)":"7B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### Random guessing

Existence: 25. Result ID: 6a31130bfbc373592c2d9ec2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"-","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

### G-LLaVA

Existence: 22.1. Result ID: 6a31130bfbc373592c2d9edf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77071e209131cea709bb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ac11e209131cea8503e.md>)

Recorded evaluation configuration: {"Model Size (Parameters)":"7B","Zero-shot manner":"true","Temperature":"0","Image Resolution":"640x640"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
