# Generalization Bound Estimation on ImageNet (val)

Source: [Generalization Bound Estimation on ImageNet (val)](<https://www.sota2.com/research/sota/generalization-bound-estimation-on-imagenet-val>)

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

Data updated: 2026-06-02T09:56:30.480Z

Evaluation of theoretical generalization bounds (e.g., Kawaguchi et al., Than et al.) and test error estimation on the validation split of the ImageNet 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 | Generalization Bound Estimation |
| Dataset (including any stated split) | ImageNet (val) |
| Metric | Test Error |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 17 of 17 |
| 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

- [Test Error](<https://www.sota2.com/research/sota/generalization-bound-estimation-on-imagenet-val.md?metric=Test%20Error>): Lower is better; unit: SCALAR.

- [Bound (Kawaguchi et al., 2022)](<https://www.sota2.com/research/sota/generalization-bound-estimation-on-imagenet-val.md?metric=Bound%20(Kawaguchi%20et%20al.%2C%202022)>): Lower is better; unit: SCALAR.

- [Bound (Than et al., 2025)](<https://www.sota2.com/research/sota/generalization-bound-estimation-on-imagenet-val.md?metric=Bound%20(Than%20et%20al.%2C%202025)>): Lower is better; unit: SCALAR.

- [Our Bound (2)](<https://www.sota2.com/research/sota/generalization-bound-estimation-on-imagenet-val.md?metric=Our%20Bound%20(2)>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Test Error | Result date | Report paper |
| --- | --- | --- | --- |
| SwinTransformer B V2 | 0.159 | Not recorded | 68fa77561e209131cea727fc |
| SwinTransformer B | 0.164 | Not recorded | 68fa77561e209131cea727fc |
| SwinTransformer T V2 | 0.179 | Not recorded | 68fa77561e209131cea727fc |
| SwinTransformer T | 0.185 | Not recorded | 68fa77561e209131cea727fc |
| ResNet152 V1 | 0.217 | Not recorded | 68fa77561e209131cea727fc |
| ResNet101 V1 | 0.226 | Not recorded | 68fa77561e209131cea727fc |
| DenseNet161 | 0.229 | Not recorded | 68fa77561e209131cea727fc |
| DenseNet201 | 0.231 | Not recorded | 68fa77561e209131cea727fc |
| ResNet50 V1 | 0.239 | Not recorded | 68fa77561e209131cea727fc |
| DenseNet169 | 0.244 | Not recorded | 68fa77561e209131cea727fc |
| DenseNet121 | 0.256 | Not recorded | 68fa77561e209131cea727fc |
| VGG19 BN | 0.258 | Not recorded | 68fa77561e209131cea727fc |
| ResNet34 V1 | 0.267 | Not recorded | 68fa77561e209131cea727fc |
| VGG19 | 0.276 | Not recorded | 68fa77561e209131cea727fc |
| VGG13 BN | 0.284 | Not recorded | 68fa77561e209131cea727fc |
| VGG13 | 0.301 | Not recorded | 68fa77561e209131cea727fc |
| ResNet18 V1 | 0.302 | Not recorded | 68fa77561e209131cea727fc |

### SwinTransformer B V2

Test Error: 0.159. Result ID: 6a1ea887fbc373592c1945a8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"SwinTransformer B V2"}. This record does not assert independent reproduction.

### SwinTransformer B

Test Error: 0.164. Result ID: 6a1ea887fbc373592c1945a6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"SwinTransformer B"}. This record does not assert independent reproduction.

### SwinTransformer T V2

Test Error: 0.179. Result ID: 6a1ea887fbc373592c1945a9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"SwinTransformer T V2"}. This record does not assert independent reproduction.

### SwinTransformer T

Test Error: 0.185. Result ID: 6a1ea887fbc373592c1945a7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"SwinTransformer T"}. This record does not assert independent reproduction.

### ResNet152 V1

Test Error: 0.217. Result ID: 6a1ea887fbc373592c1945a5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"ResNet152 V1"}. This record does not assert independent reproduction.

### ResNet101 V1

Test Error: 0.226. Result ID: 6a1ea887fbc373592c1945a4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"ResNet101 V1"}. This record does not assert independent reproduction.

### DenseNet161

Test Error: 0.229. Result ID: 6a1ea887fbc373592c1945af.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"DenseNet161"}. This record does not assert independent reproduction.

### DenseNet201

Test Error: 0.231. Result ID: 6a1ea887fbc373592c1945b1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"DenseNet201"}. This record does not assert independent reproduction.

### ResNet50 V1

Test Error: 0.239. Result ID: 6a1ea887fbc373592c1945a3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"ResNet50 V1"}. This record does not assert independent reproduction.

### DenseNet169

Test Error: 0.244. Result ID: 6a1ea887fbc373592c1945b0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"DenseNet169"}. This record does not assert independent reproduction.

### DenseNet121

Test Error: 0.256. Result ID: 6a1ea887fbc373592c1945ae.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"DenseNet121"}. This record does not assert independent reproduction.

### VGG19 BN

Test Error: 0.258. Result ID: 6a1ea887fbc373592c1945ad.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"VGG19 BN"}. This record does not assert independent reproduction.

### ResNet34 V1

Test Error: 0.267. Result ID: 6a1ea887fbc373592c1945a2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"ResNet34 V1"}. This record does not assert independent reproduction.

### VGG19

Test Error: 0.276. Result ID: 6a1ea887fbc373592c1945ac.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"VGG19"}. This record does not assert independent reproduction.

### VGG13 BN

Test Error: 0.284. Result ID: 6a1ea887fbc373592c1945ab.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"VGG13 BN"}. This record does not assert independent reproduction.

### VGG13

Test Error: 0.301. Result ID: 6a1ea887fbc373592c1945aa.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"VGG13"}. This record does not assert independent reproduction.

### ResNet18 V1

Test Error: 0.302. Result ID: 6a1ea887fbc373592c1945a1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77561e209131cea727fc.md>)

Recorded evaluation configuration: {"Model":"ResNet18 V1"}. This record does not assert independent reproduction.

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