# Visual Question Answering on GQA (Sys.)

Source: [Visual Question Answering on GQA (Sys.)](<https://www.sota2.com/research/sota/visual-question-answering-on-gqa-sys>)

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

Data updated: 2026-02-26T05:54:23.904Z

Evaluation of model performance on the system split of the GQA dataset for visual question answering capabilities.

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 | Visual Question Answering |
| Dataset (including any stated split) | GQA (Sys.) |
| Metric | Overall Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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/visual-question-answering-on-gqa-sys.md?metric=Overall%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Overall Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| MCAN-Small | 36.21 | Not recorded | 68fa6cc21e209131cea35d61 |
| RelViT | 35.48 | Not recorded | 68fa6cc21e209131cea35d61 |
| EsViT | 31.76 | Not recorded | 68fa6cc21e209131cea35d61 |
| ViT-only | 31.39 | Not recorded | 68fa6cc21e209131cea35d61 |
| MCAN-Small | 30.12 | Not recorded | 68fa6cc21e209131cea35d61 |

### MCAN-Small

Overall Accuracy: 36.21. Result ID: 698f3c683a6c314849f755c8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6cc21e209131cea35d61.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65f11e209131cea0d447.md>)

[Implementation](<https://github.com/MILVLG/mcan-vqa>)

Recorded evaluation configuration: {"Faster R-CNN bbox features":true,"Backbone":"ResNet-101"}. This record does not assert independent reproduction.

### RelViT

Overall Accuracy: 35.48. Result ID: 698f3c683a6c314849f755cc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6cc21e209131cea35d61.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cc21e209131cea35d61.md>)

Recorded evaluation configuration: {"Faster R-CNN bbox features":false,"Backbone":"PVTv2-b2"}. This record does not assert independent reproduction.

### EsViT

Overall Accuracy: 31.76. Result ID: 698f3c683a6c314849f755cb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6cc21e209131cea35d61.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cfa1e209131cea37111.md>)

[Implementation](<https://github.com/microsoft/esvit>)

Recorded evaluation configuration: {"Faster R-CNN bbox features":false,"Backbone":"PVTv2-b2"}. This record does not assert independent reproduction.

### ViT-only

Overall Accuracy: 31.39. Result ID: 698f3c683a6c314849f755ca.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6cc21e209131cea35d61.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f131e209131cea43b6a.md>)

[Implementation](<https://github.com/whai362/PVT>)

Recorded evaluation configuration: {"Faster R-CNN bbox features":false,"Backbone":"PVTv2-b2"}. This record does not assert independent reproduction.

### MCAN-Small

Overall Accuracy: 30.12. Result ID: 698f3c683a6c314849f755c9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6cc21e209131cea35d61.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65f11e209131cea0d447.md>)

[Implementation](<https://github.com/MILVLG/mcan-vqa>)

Recorded evaluation configuration: {"Faster R-CNN bbox features":false,"Backbone":"ResNet-101"}. This record does not assert independent reproduction.

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