# Noisy Test-Time Adaptation on ImageNet-S

Source: [Noisy Test-Time Adaptation on ImageNet-S](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s>)

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

Data updated: 2026-06-25T08:37:17.180Z

Evaluation of Test-Time Adaptation performance on the ImageNet-S dataset, reporting accuracy metrics across various corruption sources (iNaturalist, SUN, Texture, Places).

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 | Noisy Test-Time Adaptation |
| Dataset (including any stated split) | ImageNet-S |
| Metric | Accuracy (iNaturalist, Severity S) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| 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 (iNaturalist, Severity S)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(iNaturalist%2C%20Severity%20S)>): Higher is better; unit: PERCENT.

- [Accuracy (iNaturalist, Severity N)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(iNaturalist%2C%20Severity%20N)>): Higher is better; unit: PERCENT.

- [Accuracy (iNaturalist, Severity H)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(iNaturalist%2C%20Severity%20H)>): Higher is better; unit: PERCENT.

- [Accuracy (SUN, Severity S)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(SUN%2C%20Severity%20S)>): Higher is better; unit: PERCENT.

- [Accuracy (SUN, Severity N)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(SUN%2C%20Severity%20N)>): Higher is better; unit: PERCENT.

- [Accuracy (SUN, Severity H)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(SUN%2C%20Severity%20H)>): Higher is better; unit: PERCENT.

- [Accuracy (Texture, Severity S)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(Texture%2C%20Severity%20S)>): Higher is better; unit: PERCENT.

- [Accuracy (Texture, Severity N)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(Texture%2C%20Severity%20N)>): Higher is better; unit: PERCENT.

- [Accuracy (Texture, Severity H)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(Texture%2C%20Severity%20H)>): Higher is better; unit: PERCENT.

- [Accuracy (Places, Severity S)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(Places%2C%20Severity%20S)>): Higher is better; unit: PERCENT.

- [Accuracy (Places, Severity N)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(Places%2C%20Severity%20N)>): Higher is better; unit: PERCENT.

- [Accuracy (Places, Severity H)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Accuracy%20(Places%2C%20Severity%20H)>): Higher is better; unit: PERCENT.

- [Average Accuracy (Severity S)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Average%20Accuracy%20(Severity%20S)>): Higher is better; unit: PERCENT.

- [Average Accuracy (Severity N)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Average%20Accuracy%20(Severity%20N)>): Higher is better; unit: PERCENT.

- [Average Accuracy (Severity H)](<https://www.sota2.com/research/sota/noisy-test-time-adaptation-on-imagenet-s.md?metric=Average%20Accuracy%20(Severity%20H)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy (iNaturalist, Severity S) | Result date | Report paper |
| --- | --- | --- | --- |
| DDE | 43.23 | Not recorded | 6a3cb624206542251d7ca61f |
| DMN | 42.69 | Not recorded | 6a3cb624206542251d7ca61f |
| OODD | 42.41 | Not recorded | 6a3cb624206542251d7ca61f |
| AdaND | 40.97 | Not recorded | 6a3cb624206542251d7ca61f |
| AdaNeg | 39.9 | Not recorded | 6a3cb624206542251d7ca61f |
| ZS-CLIP | 34.17 | Not recorded | 6a3cb624206542251d7ca61f |
| TPT | 32.16 | Not recorded | 6a3cb624206542251d7ca61f |
| Tent | 30.46 | Not recorded | 6a3cb624206542251d7ca61f |

### DDE

Accuracy (iNaturalist, Severity S): 43.23. Result ID: 6a3ce8abfbc373592c527ee1.

[Reporting paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

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

### DMN

Accuracy (iNaturalist, Severity S): 42.69. Result ID: 6a3ce8abfbc373592c527edd.

[Reporting paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72b31e209131cea58047.md>)

[Implementation](<https://github.com/YBZh/DMN>)

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

### OODD

Accuracy (iNaturalist, Severity S): 42.41. Result ID: 6a3ce8abfbc373592c527edf.

[Reporting paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77631e209131cea72ea6.md>)

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

### AdaND

Accuracy (iNaturalist, Severity S): 40.97. Result ID: 6a3ce8abfbc373592c527ee0.

[Reporting paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77e11e209131cea75c33.md>)

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

### AdaNeg

Accuracy (iNaturalist, Severity S): 39.9. Result ID: 6a3ce8abfbc373592c527ede.

[Reporting paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75711e209131cea67f50.md>)

[Implementation](<https://github.com/YBZh/OpenOOD-VLM>)

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

### ZS-CLIP

Accuracy (iNaturalist, Severity S): 34.17. Result ID: 6a3ce8abfbc373592c527eda.

[Reporting paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

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

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

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

### TPT

Accuracy (iNaturalist, Severity S): 32.16. Result ID: 6a3ce8abfbc373592c527edc.

[Reporting paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d691e209131cea399f9.md>)

[Implementation](<https://tinyurl.com/yr3zmhma>)

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

### Tent

Accuracy (iNaturalist, Severity S): 30.46. Result ID: 6a3ce8abfbc373592c527edb.

[Reporting paper](<https://www.sota2.com/research/paper/6a3cb624206542251d7ca61f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69941e209131cea22d2b.md>)

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

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