# Fine-grained Visual Question Answering on HRBench-8K

Source: [Fine-grained Visual Question Answering on HRBench-8K](<https://www.sota2.com/research/sota/fine-grained-visual-question-answering-on-hrbench-8k>)

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

Data updated: 2026-02-26T05:54:23.108Z

Evaluation of model performance on the HRBench-8K dataset for Fine-grained Visual Question Answering, reporting metrics including Overall Accuracy, FSP, FCP, and FP.

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 | Fine-grained Visual Question Answering |
| Dataset (including any stated split) | HRBench-8K |
| Metric | Overall Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 28 of 28 |
| 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/fine-grained-visual-question-answering-on-hrbench-8k.md?metric=Overall%20Accuracy>): Higher is better; unit: PERCENT.

- [FSP](<https://www.sota2.com/research/sota/fine-grained-visual-question-answering-on-hrbench-8k.md?metric=FSP>): Higher is better; unit: PERCENT.

- [FCP](<https://www.sota2.com/research/sota/fine-grained-visual-question-answering-on-hrbench-8k.md?metric=FCP>): Higher is better; unit: PERCENT.

- [FP](<https://www.sota2.com/research/sota/fine-grained-visual-question-answering-on-hrbench-8k.md?metric=FP>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Overall Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| FOCUS | 69.63 | Not recorded | 68fa79b41e209131cea7efbe |
| ZoomEye | 69 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 68.88 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 68.75 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 68.38 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 67.63 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 66.88 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 66.88 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 65.75 | Not recorded | 68fa79b41e209131cea7efbe |
| ZoomEye | 65.63 | Not recorded | 68fa79b41e209131cea7efbe |
| ZoomEye | 59.63 | Not recorded | 68fa79b41e209131cea7efbe |
| ZoomEye | 59 | Not recorded | 68fa79b41e209131cea7efbe |
| ZoomEye | 49 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 45 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 45 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 44.13 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 44.13 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 42.38 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 42.15 | Not recorded | 68fa79b41e209131cea7efbe |
| ZoomEye | 42.13 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 40.63 | Not recorded | 68fa79b41e209131cea7efbe |
| FOCUS | 40.63 | Not recorded | 68fa79b41e209131cea7efbe |
| DC2 | 39.5 | Not recorded | 68fa79b41e209131cea7efbe |
| ViCrop | 39.38 | Not recorded | 68fa79b41e209131cea7efbe |
| ViCrop | 38.38 | Not recorded | 68fa79b41e209131cea7efbe |
| ZoomEye | 33.88 | Not recorded | 68fa79b41e209131cea7efbe |
| SEAL | 33.5 | Not recorded | 68fa79b41e209131cea7efbe |
| ZoomEye | 32.88 | Not recorded | 68fa79b41e209131cea7efbe |

### FOCUS

Overall Accuracy: 69.63. Result ID: 698b59ef7a967eddb76ddd4a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Steps":8,"Overrun":true}. This record does not assert independent reproduction.

### ZoomEye

Overall Accuracy: 69. Result ID: 698b59ef7a967eddb76ddd36.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af51e209131cea85f9c.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Depth":"5","Number of crops":"4"}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 68.88. Result ID: 698b59ef7a967eddb76ddd49.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Steps":6,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 68.75. Result ID: 698b59ef7a967eddb76ddd45.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Steps":1,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 68.38. Result ID: 698b59ef7a967eddb76ddd48.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Steps":4,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 67.63. Result ID: 698b59ef7a967eddb76ddd46.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Steps":2,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 66.88. Result ID: 698b59ef7a967eddb76ddd44.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Steps":2,"Overrun":false}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 66.88. Result ID: 698b59ef7a967eddb76ddd47.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Steps":3,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 65.75. Result ID: 698b59ef7a967eddb76ddd43.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Steps":1,"Overrun":false}. This record does not assert independent reproduction.

### ZoomEye

Overall Accuracy: 65.63. Result ID: 698b59ef7a967eddb76ddd35.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af51e209131cea85f9c.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Depth":"2","Number of crops":"4"}. This record does not assert independent reproduction.

### ZoomEye

Overall Accuracy: 59.63. Result ID: 698b59ef7a967eddb76ddd34.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af51e209131cea85f9c.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Depth":"1","Number of crops":"4"}. This record does not assert independent reproduction.

### ZoomEye

Overall Accuracy: 59. Result ID: 698b59ef7a967eddb76ddd33.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af51e209131cea85f9c.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV-7B","Depth":"1","Number of crops":"2"}. This record does not assert independent reproduction.

### ZoomEye

Overall Accuracy: 49. Result ID: 698b59ef7a967eddb76ddd32.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af51e209131cea85f9c.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Depth":"5","Number of crops":"4"}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 45. Result ID: 698b59ef7a967eddb76ddd40.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Steps":4,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 45. Result ID: 698b59ef7a967eddb76ddd41.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Steps":6,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 44.13. Result ID: 698b59ef7a967eddb76ddd3f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Steps":3,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 44.13. Result ID: 698b59ef7a967eddb76ddd42.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Steps":8,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 42.38. Result ID: 698b59ef7a967eddb76ddd3d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Steps":1,"Overrun":true}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 42.15. Result ID: 698b59ef7a967eddb76ddd3e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Steps":2,"Overrun":true}. This record does not assert independent reproduction.

### ZoomEye

Overall Accuracy: 42.13. Result ID: 698b59ef7a967eddb76ddd31.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af51e209131cea85f9c.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Depth":"2","Number of crops":"4"}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 40.63. Result ID: 698b59ef7a967eddb76ddd3b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Steps":1,"Overrun":false}. This record does not assert independent reproduction.

### FOCUS

Overall Accuracy: 40.63. Result ID: 698b59ef7a967eddb76ddd3c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Implementation](<https://focus-mllm-vqa.github.io/>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Steps":2,"Overrun":false}. This record does not assert independent reproduction.

### DC2

Overall Accuracy: 39.5. Result ID: 698b59ef7a967eddb76ddd39.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74a91e209131cea63071.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-v.1.5-7B"}. This record does not assert independent reproduction.

### ViCrop

Overall Accuracy: 39.38. Result ID: 698b59ef7a967eddb76ddd37.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa771a1e209131cea71032.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Mode":"rel-att-high"}. This record does not assert independent reproduction.

### ViCrop

Overall Accuracy: 38.38. Result ID: 698b59ef7a967eddb76ddd38.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa771a1e209131cea71032.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Mode":"grad-att-high"}. This record does not assert independent reproduction.

### ZoomEye

Overall Accuracy: 33.88. Result ID: 698b59ef7a967eddb76ddd2f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af51e209131cea85f9c.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Depth":"1","Number of crops":"2"}. This record does not assert independent reproduction.

### SEAL

Overall Accuracy: 33.5. Result ID: 698b59ef7a967eddb76ddd3a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71ac1e209131cea52781.md>)

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

### ZoomEye

Overall Accuracy: 32.88. Result ID: 698b59ef7a967eddb76ddd30.

[Reporting paper](<https://www.sota2.com/research/paper/68fa79b41e209131cea7efbe.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7af51e209131cea85f9c.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-1.5-7B","Depth":"1","Number of crops":"4"}. This record does not assert independent reproduction.

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