# Visual Question Answering and Grounding on VizWiz (test)

Source: [Visual Question Answering and Grounding on VizWiz (test)](<https://www.sota2.com/research/sota/visual-question-answering-and-grounding-on-vizwiz-test>)

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

Data updated: 2026-04-21T10:35:43.048Z

Evaluation of Visual Question Answering and Grounding performance on the test split of the VizWiz dataset, measured using metrics such as CLIPScore, TIFA, and VQA-G scores.

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 and Grounding |
| Dataset (including any stated split) | VizWiz (test) |
| Metric | CLIPScore |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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

- [CLIPScore](<https://www.sota2.com/research/sota/visual-question-answering-and-grounding-on-vizwiz-test.md?metric=CLIPScore>): Higher is better; unit: SCALAR.

- [TIFA](<https://www.sota2.com/research/sota/visual-question-answering-and-grounding-on-vizwiz-test.md?metric=TIFA>): Higher is better; unit: SCALAR.

- [VQA Score](<https://www.sota2.com/research/sota/visual-question-answering-and-grounding-on-vizwiz-test.md?metric=VQA%20Score>): Higher is better; unit: SCALAR.

- [mIoU](<https://www.sota2.com/research/sota/visual-question-answering-and-grounding-on-vizwiz-test.md?metric=mIoU>): Higher is better; unit: PERCENT.

- [Acc@0.5IoU](<https://www.sota2.com/research/sota/visual-question-answering-and-grounding-on-vizwiz-test.md?metric=Acc%400.5IoU>): Higher is better; unit: PERCENT.

- [Average VQA-G Score](<https://www.sota2.com/research/sota/visual-question-answering-and-grounding-on-vizwiz-test.md?metric=Average%20VQA-G%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | CLIPScore | Result date | Report paper |
| --- | --- | --- | --- |
| AutoVQA-G | 0.757 | Not recorded | 69e704a67d0172140ec7ee04 |
| GPT-4o | 0.754 | Not recorded | 69e704a67d0172140ec7ee04 |
| GPT-4o | 0.753 | Not recorded | 69e704a67d0172140ec7ee04 |
| Gemini | 0.745 | Not recorded | 69e704a67d0172140ec7ee04 |
| Gemini | 0.745 | Not recorded | 69e704a67d0172140ec7ee04 |
| Human Annotation | 0.724 | Not recorded | 69e704a67d0172140ec7ee04 |

### AutoVQA-G

CLIPScore: 0.757. Result ID: 69e7516747f34ce924ce442c.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee04.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee04.md>)

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

### GPT-4o

CLIPScore: 0.754. Result ID: 69e7516747f34ce924ce4429.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee04.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75681e209131cea67a98.md>)

Recorded evaluation configuration: {"chain-of-thought":"true"}. This record does not assert independent reproduction.

### GPT-4o

CLIPScore: 0.753. Result ID: 69e7516747f34ce924ce4428.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee04.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75681e209131cea67a98.md>)

Recorded evaluation configuration: {"zero-shot":"true","tool-assisted":"true"}. This record does not assert independent reproduction.

### Gemini

CLIPScore: 0.745. Result ID: 69e7516747f34ce924ce442a.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee04.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c311e209131cea8cef5.md>)

Recorded evaluation configuration: {"zero-shot":"true"}. This record does not assert independent reproduction.

### Gemini

CLIPScore: 0.745. Result ID: 69e7516747f34ce924ce442b.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee04.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c311e209131cea8cef5.md>)

Recorded evaluation configuration: {"chain-of-thought":"true"}. This record does not assert independent reproduction.

### Human Annotation

CLIPScore: 0.724. Result ID: 69e7516747f34ce924ce4427.

[Reporting paper](<https://www.sota2.com/research/paper/69e704a67d0172140ec7ee04.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.