# Image Classification on Unseen Tasks (SUN397, Cars, RESISC45, DTD) (test)

Source: [Image Classification on Unseen Tasks (SUN397, Cars, RESISC45, DTD) (test)](<https://www.sota2.com/research/sota/image-classification-on-unseen-tasks-sun397-cars-resisc45-dtd-test>)

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

Data updated: 2026-03-05T09:00:02.765Z

Aggregate evaluation of Image Classification performance across the test splits of four specific datasets identified as 'Unseen Tasks': SUN397, Cars, RESISC45, and DTD.

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 | Image Classification |
| Dataset (including any stated split) | Unseen Tasks (SUN397, Cars, RESISC45, DTD) (test) |
| Metric | SUN397 Accuracy |
| Unit | PERCENT |
| 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

- [SUN397 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-unseen-tasks-sun397-cars-resisc45-dtd-test.md?metric=SUN397%20Accuracy>): Higher is better; unit: PERCENT.

- [Cars Accuracy](<https://www.sota2.com/research/sota/image-classification-on-unseen-tasks-sun397-cars-resisc45-dtd-test.md?metric=Cars%20Accuracy>): Higher is better; unit: PERCENT.

- [RESISC45 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-unseen-tasks-sun397-cars-resisc45-dtd-test.md?metric=RESISC45%20Accuracy>): Higher is better; unit: PERCENT.

- [DTD Accuracy](<https://www.sota2.com/research/sota/image-classification-on-unseen-tasks-sun397-cars-resisc45-dtd-test.md?metric=DTD%20Accuracy>): Higher is better; unit: PERCENT.

- [Average Accuracy](<https://www.sota2.com/research/sota/image-classification-on-unseen-tasks-sun397-cars-resisc45-dtd-test.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | SUN397 Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Pretrained | 62.46 | Not recorded | 69a90e142d603e6de3e408af |
| Ties-Merging | 60.19 | Not recorded | 69a90e142d603e6de3e408af |
| BD-Merging | 60.09 | Not recorded | 69a90e142d603e6de3e408af |
| Twin-Merging | 59.72 | Not recorded | 69a90e142d603e6de3e408af |
| Task Arithmetic | 56.41 | Not recorded | 69a90e142d603e6de3e408af |
| AdaMerging | 56.36 | Not recorded | 69a90e142d603e6de3e408af |

### Pretrained

SUN397 Accuracy: 62.46. Result ID: 69a945fe6dbb7075b952e3ac.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e142d603e6de3e408af.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32"}. This record does not assert independent reproduction.

### Ties-Merging

SUN397 Accuracy: 60.19. Result ID: 69a945fe6dbb7075b952e3ae.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e142d603e6de3e408af.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71091e209131cea4ecd8.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Weights":"Layer-wise"}. This record does not assert independent reproduction.

### BD-Merging

SUN397 Accuracy: 60.09. Result ID: 69a945fe6dbb7075b952e3b1.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e142d603e6de3e408af.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69a90e142d603e6de3e408af.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Weights":"Layer-wise"}. This record does not assert independent reproduction.

### Twin-Merging

SUN397 Accuracy: 59.72. Result ID: 69a945fe6dbb7075b952e3b0.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e142d603e6de3e408af.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75381e209131cea66883.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Weights":"Layer-wise"}. This record does not assert independent reproduction.

### Task Arithmetic

SUN397 Accuracy: 56.41. Result ID: 69a945fe6dbb7075b952e3ad.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e142d603e6de3e408af.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ef01e209131cea43012.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Weights":"Layer-wise"}. This record does not assert independent reproduction.

### AdaMerging

SUN397 Accuracy: 56.36. Result ID: 69a945fe6dbb7075b952e3af.

[Reporting paper](<https://www.sota2.com/research/paper/69a90e142d603e6de3e408af.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73691e209131cea5c535.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/32","Weights":"Layer-wise"}. This record does not assert independent reproduction.

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