# Image Classification on CIFAR-10 (test) (Dirichlet & Pathological Accuracy)

Source: [Image Classification on CIFAR-10 (test) (Dirichlet & Pathological Accuracy)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-dirichlet-and-pathological-accuracy>)

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

Data updated: 2026-02-27T06:04:02.372Z

Evaluation of Image Classification performance on the CIFAR-10 test set, reporting accuracy metrics derived from Dirichlet sampling and pathological perturbation tests.

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) | CIFAR-10 (test) |
| Metric | Accuracy (Dirichlet α=0.1) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| 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 (Dirichlet α=0.1)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-dirichlet-and-pathological-accuracy.md?metric=Accuracy%20(Dirichlet%20%CE%B1%3D0.1)>): Higher is better; unit: PERCENT.

- [Accuracy (Dirichlet α=0.3)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-dirichlet-and-pathological-accuracy.md?metric=Accuracy%20(Dirichlet%20%CE%B1%3D0.3)>): Higher is better; unit: PERCENT.

- [Accuracy (Pathological c=2)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-dirichlet-and-pathological-accuracy.md?metric=Accuracy%20(Pathological%20c%3D2)>): Higher is better; unit: PERCENT.

- [Accuracy (Pathological c=5)](<https://www.sota2.com/research/sota/image-classification-on-cifar-10-test-dirichlet-and-pathological-accuracy.md?metric=Accuracy%20(Pathological%20c%3D5)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy (Dirichlet α=0.1) | Result date | Report paper |
| --- | --- | --- | --- |
| DFedPGP | 88.85 | Not recorded | 68fa5bd71e209131ce9cc033 |
| FedRep | 88.78 | Not recorded | 68fa5bd71e209131ce9cc033 |
| FedPer | 88.57 | Not recorded | 68fa5bd71e209131ce9cc033 |
| FedBABU | 87.79 | Not recorded | 68fa5bd71e209131ce9cc033 |
| Dis-PFL | 87.77 | Not recorded | 68fa5bd71e209131ce9cc033 |
| OSGP | 87.39 | Not recorded | 68fa5bd71e209131ce9cc033 |
| DFedAvgM | 86.94 | Not recorded | 68fa5bd71e209131ce9cc033 |
| FedAvg | 84.17 | Not recorded | 68fa5bd71e209131ce9cc033 |
| Ditto | 80.22 | Not recorded | 68fa5bd71e209131ce9cc033 |
| Local | 78.96 | Not recorded | 68fa5bd71e209131ce9cc033 |

### DFedPGP

Accuracy (Dirichlet α=0.1): 88.85. Result ID: 69a133bb35dfcabb00e5786e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD","Local training epochs (shared part)":"5","Local training epochs (personal part)":"1","PFL Layer Partitioning":"classifier as personal"}. This record does not assert independent reproduction.

### FedRep

Accuracy (Dirichlet α=0.1): 88.78. Result ID: 69a133bb35dfcabb00e57868.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa5a3d1e209131ce9be169.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD","PFL Layer Partitioning":"classifier as personal"}. This record does not assert independent reproduction.

### FedPer

Accuracy (Dirichlet α=0.1): 88.57. Result ID: 69a133bb35dfcabb00e57867.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66cb1e209131cea12528.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD","PFL Layer Partitioning":"classifier as personal"}. This record does not assert independent reproduction.

### FedBABU

Accuracy (Dirichlet α=0.1): 87.79. Result ID: 69a133bb35dfcabb00e57869.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bfa1e209131cea31630.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD","PFL Layer Partitioning":"classifier as personal"}. This record does not assert independent reproduction.

### Dis-PFL

Accuracy (Dirichlet α=0.1): 87.77. Result ID: 69a133bb35dfcabb00e5786d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ca31e209131cea3522c.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD"}. This record does not assert independent reproduction.

### OSGP

Accuracy (Dirichlet α=0.1): 87.39. Result ID: 69a133bb35dfcabb00e5786c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa56dd1e209131ce9a0808.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD"}. This record does not assert independent reproduction.

### DFedAvgM

Accuracy (Dirichlet α=0.1): 86.94. Result ID: 69a133bb35dfcabb00e5786b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69cb1e209131cea24199.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD"}. This record does not assert independent reproduction.

### FedAvg

Accuracy (Dirichlet α=0.1): 84.17. Result ID: 69a133bb35dfcabb00e57866.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e761e209131cea40072.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Client Sampling Ratio":"0.1","Batch Size":"128","Optimizer":"SGD"}. This record does not assert independent reproduction.

### Ditto

Accuracy (Dirichlet α=0.1): 80.22. Result ID: 69a133bb35dfcabb00e5786a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a2d1e209131cea269df.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD"}. This record does not assert independent reproduction.

### Local

Accuracy (Dirichlet α=0.1): 78.96. Result ID: 69a133bb35dfcabb00e57865.

[Reporting paper](<https://www.sota2.com/research/paper/68fa5bd71e209131ce9cc033.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-18","Normalization":"Group Normalization","Communication Rounds":"500","Number of Clients":"100","Batch Size":"128","Optimizer":"SGD"}. This record does not assert independent reproduction.

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