# Multi-task image classification on Multi-Fashion MNIST (test)

Source: [Multi-task image classification on Multi-Fashion MNIST (test)](<https://www.sota2.com/research/sota/multi-task-image-classification-on-multi-fashion-mnist-test>)

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

Data updated: 2026-02-26T05:34:34.784Z

Evaluation of Multi-task Image Classification performance on the test split of the Multi-Fashion MNIST dataset.

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 | Multi-task image classification |
| Dataset (including any stated split) | Multi-Fashion MNIST (test) |
| Metric | Accuracy 1 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 7 of 7 |
| 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

- [Accuracy 1](<https://www.sota2.com/research/sota/multi-task-image-classification-on-multi-fashion-mnist-test.md?metric=Accuracy%201>): Higher is better; unit: PERCENT.

- [Accuracy 2](<https://www.sota2.com/research/sota/multi-task-image-classification-on-multi-fashion-mnist-test.md?metric=Accuracy%202>): Higher is better; unit: PERCENT.

- [Experts](<https://www.sota2.com/research/sota/multi-task-image-classification-on-multi-fashion-mnist-test.md?metric=Experts>): Higher is better; unit: SCALAR.

- [Ensemble Accuracy](<https://www.sota2.com/research/sota/multi-task-image-classification-on-multi-fashion-mnist-test.md?metric=Ensemble%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy 1 | Result date | Report paper |
| --- | --- | --- | --- |
| DSelect-k | 83.78 | Not recorded | 68fa58d31e209131ce9b185c |
| DSelect-k | 83.69 | Not recorded | 68fa58d31e209131ce9b185c |
| Gumbel Softmax | 83.66 | Not recorded | 68fa58d31e209131ce9b185c |
| Top-k | 83.66 | Not recorded | 68fa58d31e209131ce9b185c |
| Softmax MoE | 83.48 | Not recorded | 68fa58d31e209131ce9b185c |
| Top-k | 83.44 | Not recorded | 68fa58d31e209131ce9b185c |
| Shared Bottom | 82.05 | Not recorded | 68fa58d31e209131ce9b185c |

### DSelect-k

Accuracy 1: 83.78. Result ID: 698ba4fe7a967eddb7704a22.

[Reporting paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

[Implementation](<https://github.com/google-research/google-research/tree/master/dselect_k_moe>)

Recorded evaluation configuration: {"Gate selection type":"Static"}. This record does not assert independent reproduction.

### DSelect-k

Accuracy 1: 83.69. Result ID: 698ba4fe7a967eddb7704a25.

[Reporting paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

[Implementation](<https://github.com/google-research/google-research/tree/master/dselect_k_moe>)

Recorded evaluation configuration: {"Gate selection type":"Per-example"}. This record does not assert independent reproduction.

### Gumbel Softmax

Accuracy 1: 83.66. Result ID: 698ba4fe7a967eddb7704a24.

[Reporting paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68c41e209131cea1e3b8.md>)

[Implementation](<https://cs-people.bu.edu/sunxm/AdaShare/project.html>)

Recorded evaluation configuration: {"Gate selection type":"Static"}. This record does not assert independent reproduction.

### Top-k

Accuracy 1: 83.66. Result ID: 698ba4fe7a967eddb7704a26.

[Reporting paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63171e209131ce9fc941.md>)

Recorded evaluation configuration: {"Gate selection type":"Per-example"}. This record does not assert independent reproduction.

### Softmax MoE

Accuracy 1: 83.48. Result ID: 698ba4fe7a967eddb7704a27.

[Reporting paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

Recorded evaluation configuration: {"Gate selection type":"Baseline"}. This record does not assert independent reproduction.

### Top-k

Accuracy 1: 83.44. Result ID: 698ba4fe7a967eddb7704a23.

[Reporting paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63171e209131ce9fc941.md>)

Recorded evaluation configuration: {"Gate selection type":"Static"}. This record does not assert independent reproduction.

### Shared Bottom

Accuracy 1: 82.05. Result ID: 698ba4fe7a967eddb7704a28.

[Reporting paper](<https://www.sota2.com/research/paper/68fa58d31e209131ce9b185c.md>)

Recorded evaluation configuration: {"Gate selection type":"Baseline"}. This record does not assert independent reproduction.

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