# High-resolution Visual Question Answering on HR-Bench 4K

Source: [High-resolution Visual Question Answering on HR-Bench 4K](<https://www.sota2.com/research/sota/high-resolution-visual-question-answering-on-hr-bench-4k>)

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

Data updated: 2026-05-05T06:54:48.142Z

Evaluation of model performance on the HR-Bench 4K dataset specifically focusing on High-resolution Visual Question Answering, reporting scores across Overall, Single-image, and Cross-image query types.

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 | High-resolution Visual Question Answering |
| Dataset (including any stated split) | HR-Bench 4K |
| Metric | Overall Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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

- [Overall Accuracy](<https://www.sota2.com/research/sota/high-resolution-visual-question-answering-on-hr-bench-4k.md?metric=Overall%20Accuracy>): Higher is better; unit: PERCENT.

- [Single-image Accuracy](<https://www.sota2.com/research/sota/high-resolution-visual-question-answering-on-hr-bench-4k.md?metric=Single-image%20Accuracy>): Higher is better; unit: PERCENT.

- [Cross-image Accuracy](<https://www.sota2.com/research/sota/high-resolution-visual-question-answering-on-hr-bench-4k.md?metric=Cross-image%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Overall Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen3-VL-32B | 82.1 | Not recorded | 69f979fbc0f56ed843089c95 |
| PFlowNet | 80.4 | Not recorded | 69f979fbc0f56ed843089c95 |
| Qwen2.5-VL-72B | 79.4 | Not recorded | 69f979fbc0f56ed843089c95 |
| DeepEyesV2 | 77.9 | Not recorded | 69f979fbc0f56ed843089c95 |
| TreeVGR | 77.1 | Not recorded | 69f979fbc0f56ed843089c95 |
| Thyme | 77 | Not recorded | 69f979fbc0f56ed843089c95 |
| InternVL3-38B | 76.3 | Not recorded | 69f979fbc0f56ed843089c95 |
| DeepEyes | 75.1 | Not recorded | 69f979fbc0f56ed843089c95 |
| Qwen2.5-VL-32B | 74.8 | Not recorded | 69f979fbc0f56ed843089c95 |
| Qwen3-VL-4B | 73.5 | Not recorded | 69f979fbc0f56ed843089c95 |
| PixelReasoner | 72.9 | Not recorded | 69f979fbc0f56ed843089c95 |
| Qwen3-VL-8B | 72.4 | Not recorded | 69f979fbc0f56ed843089c95 |
| Qwen2.5-VL-7B | 72.1 | Not recorded | 69f979fbc0f56ed843089c95 |
| InternVL3-8B | 70.8 | Not recorded | 69f979fbc0f56ed843089c95 |
| LLaVA-OV-72B | 66.3 | Not recorded | 69f979fbc0f56ed843089c95 |
| GPT-4o-1120 | 59 | Not recorded | 69f979fbc0f56ed843089c95 |

### Qwen3-VL-32B

Overall Accuracy: 82.1. Result ID: 69f993c347f34ce924dcc05d.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### PFlowNet

Overall Accuracy: 80.4. Result ID: 69f993c347f34ce924dcc063.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Qwen2.5-VL-72B

Overall Accuracy: 79.4. Result ID: 69f993c347f34ce924dcc05a.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### DeepEyesV2

Overall Accuracy: 77.9. Result ID: 69f993c347f34ce924dcc060.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

[Method origin paper](<https://www.sota2.com/research/paper/691171652828b7cc6937d492.md>)

[Implementation](<https://github.com/Visual-Agent/DeepEyes>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### TreeVGR

Overall Accuracy: 77.1. Result ID: 69f993c347f34ce924dcc062.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79fd1e209131cea807a5.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Thyme

Overall Accuracy: 77. Result ID: 69f993c347f34ce924dcc061.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7aa21e209131cea84434.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### InternVL3-38B

Overall Accuracy: 76.3. Result ID: 69f993c347f34ce924dcc057.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### DeepEyes

Overall Accuracy: 75.1. Result ID: 69f993c347f34ce924dcc05f.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a75e.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Qwen2.5-VL-32B

Overall Accuracy: 74.8. Result ID: 69f993c347f34ce924dcc059.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Qwen3-VL-4B

Overall Accuracy: 73.5. Result ID: 69f993c347f34ce924dcc05b.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### PixelReasoner

Overall Accuracy: 72.9. Result ID: 69f993c347f34ce924dcc05e.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c71e209131cea7a774.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Qwen3-VL-8B

Overall Accuracy: 72.4. Result ID: 69f993c347f34ce924dcc05c.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Qwen2.5-VL-7B

Overall Accuracy: 72.1. Result ID: 69f993c347f34ce924dcc058.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### InternVL3-8B

Overall Accuracy: 70.8. Result ID: 69f993c347f34ce924dcc056.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### LLaVA-OV-72B

Overall Accuracy: 66.3. Result ID: 69f993c347f34ce924dcc055.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

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

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### GPT-4o-1120

Overall Accuracy: 59. Result ID: 69f993c347f34ce924dcc054.

[Reporting paper](<https://www.sota2.com/research/paper/69f979fbc0f56ed843089c95.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

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