# Vision-Language Understanding on Winoground

Source: [Vision-Language Understanding on Winoground](<https://www.sota2.com/research/sota/vision-language-understanding-on-winoground>)

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

Data updated: 2026-04-21T10:45:48.472Z

Evaluation of model performance on the Winoground dataset focusing on general Vision-Language Understanding capabilities, measured by Text Accuracy, Image Accuracy, and Group Accuracy.

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 | Vision-Language Understanding |
| Dataset (including any stated split) | Winoground |
| Metric | Text 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

- [Text Accuracy](<https://www.sota2.com/research/sota/vision-language-understanding-on-winoground.md?metric=Text%20Accuracy>): Higher is better; unit: PERCENT.

- [Image Accuracy](<https://www.sota2.com/research/sota/vision-language-understanding-on-winoground.md?metric=Image%20Accuracy>): Higher is better; unit: PERCENT.

- [Group Accuracy](<https://www.sota2.com/research/sota/vision-language-understanding-on-winoground.md?metric=Group%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Text Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| CROCScore | 61.5 | Not recorded | 68fa789f1e209131cea796a5 |
| VQAScore | 60.5 | Not recorded | 68fa789f1e209131cea796a5 |
| CROCScoreplain | 58.7 | Not recorded | 68fa789f1e209131cea796a5 |
| BLIP2-ITM | 42.8 | Not recorded | 68fa789f1e209131cea796a5 |
| AlignScore | 35.3 | Not recorded | 68fa789f1e209131cea796a5 |

### CROCScore

Text Accuracy: 61.5. Result ID: 69e754d647f34ce924ce5b25.

[Reporting paper](<https://www.sota2.com/research/paper/68fa789f1e209131cea796a5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa789f1e209131cea796a5.md>)

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

### VQAScore

Text Accuracy: 60.5. Result ID: 69e754d647f34ce924ce5b23.

[Reporting paper](<https://www.sota2.com/research/paper/68fa789f1e209131cea796a5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73a81e209131cea5dbb3.md>)

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

### CROCScoreplain

Text Accuracy: 58.7. Result ID: 69e754d647f34ce924ce5b24.

[Reporting paper](<https://www.sota2.com/research/paper/68fa789f1e209131cea796a5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa789f1e209131cea796a5.md>)

Recorded evaluation configuration: {"fine-tuned":false}. This record does not assert independent reproduction.

### BLIP2-ITM

Text Accuracy: 42.8. Result ID: 69e754d647f34ce924ce5b22.

[Reporting paper](<https://www.sota2.com/research/paper/68fa789f1e209131cea796a5.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: Not recorded. This record does not assert independent reproduction.

### AlignScore

Text Accuracy: 35.3. Result ID: 69e754d647f34ce924ce5b21.

[Reporting paper](<https://www.sota2.com/research/paper/68fa789f1e209131cea796a5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa757a1e209131cea683c9.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.