# Multilabel Image Classification on Imagenet 21k v1 (test)

Source: [Multilabel Image Classification on Imagenet 21k v1 (test)](<https://www.sota2.com/research/sota/multilabel-image-classification-on-imagenet-21k-v1-test>)

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

Data updated: 2026-02-26T06:09:16.171Z

Evaluation of Multilabel Image Classification performance on the test split of the Imagenet 21k v1 dataset, measured using metrics such as gAP and NLL.

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 | Multilabel Image Classification |
| Dataset (including any stated split) | Imagenet 21k v1 (test) |
| Metric | gAP |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| 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

- [gAP](<https://www.sota2.com/research/sota/multilabel-image-classification-on-imagenet-21k-v1-test.md?metric=gAP>): Higher is better; unit: PERCENT.

- [NLL](<https://www.sota2.com/research/sota/multilabel-image-classification-on-imagenet-21k-v1-test.md?metric=NLL>): Lower is better; unit: SCALAR.

## Recorded results

| Method | gAP | Result date | Report paper |
| --- | --- | --- | --- |
| Heteroscedastic PE | 47 | Not recorded | 68fa6a061e209131cea258a5 |
| Heteroscedastic Diag | 46.8 | Not recorded | 68fa6a061e209131cea258a5 |
| Homoscedastic | 45.9 | Not recorded | 68fa6a061e209131cea258a5 |
| Heteroscedastic Diag | 45.9 | Not recorded | 68fa6a061e209131cea258a5 |

### Heteroscedastic PE

gAP: 47. Result ID: 698dff7bffc3901de09a0840.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258a5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258a5.md>)

[Implementation](<https://github.com/google/uncertainty-baselines/tree/master/baselines/imagenet>)

Recorded evaluation configuration: {"Backbone":"ResNet-152","Temperature (tau)":"0.15","Covariance Matrix Rank":"50"}. This record does not assert independent reproduction.

### Heteroscedastic Diag

gAP: 46.8. Result ID: 698dff7bffc3901de09a083f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258a5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68b81e209131cea1dbf5.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-152","Temperature (tau)":"0.15"}. This record does not assert independent reproduction.

### Homoscedastic

gAP: 45.9. Result ID: 698dff7bffc3901de09a083d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258a5.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-152"}. This record does not assert independent reproduction.

### Heteroscedastic Diag

gAP: 45.9. Result ID: 698dff7bffc3901de09a083e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258a5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68b81e209131cea1dbf5.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-152","Temperature (tau)":"1.0"}. This record does not assert independent reproduction.

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