# Image Classification on ImageNet 3-hop split

Source: [Image Classification on ImageNet 3-hop split](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split>)

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

Data updated: 2026-02-26T05:45:20.749Z

Evaluation of standard Image Classification performance on the ImageNet 3-hop split dataset, utilizing Flat Hit@K and Hierarchical Precision metrics.

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) | ImageNet 3-hop split |
| Metric | Flat Hit@1 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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

- [Flat Hit@1](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Flat%20Hit%401>): Higher is better; unit: PERCENT.

- [Flat Hit@2](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Flat%20Hit%402>): Higher is better; unit: PERCENT.

- [Flat Hit@5](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Flat%20Hit%405>): Higher is better; unit: PERCENT.

- [Flat Hit@10](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Flat%20Hit%4010>): Higher is better; unit: PERCENT.

- [Flat Hit@20](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Flat%20Hit%4020>): Higher is better; unit: PERCENT.

- [Hierarchical Precision@2](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Hierarchical%20Precision%402>): Higher is better; unit: PERCENT.

- [Hierarchical Precision@5](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Hierarchical%20Precision%405>): Higher is better; unit: PERCENT.

- [Hierarchical Precision@10](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Hierarchical%20Precision%4010>): Higher is better; unit: PERCENT.

- [Hierarchical Precision@20](<https://www.sota2.com/research/sota/image-classification-on-imagenet-3-hop-split.md?metric=Hierarchical%20Precision%4020>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Flat Hit@1 | Result date | Report paper |
| --- | --- | --- | --- |
| DGP | 6.3 | Not recorded | 68fa6d491e209131cea38c51 |
| ZSL-KG | 6.3 | Not recorded | 68fa6d491e209131cea38c51 |
| SGCN | 6 | Not recorded | 68fa6d491e209131cea38c51 |
| GCNZ | 4.1 | Not recorded | 68fa6d491e209131cea38c51 |
| ZSL-KG | 3.4 | Not recorded | 68fa6d491e209131cea38c51 |
| SGCN | 3.2 | Not recorded | 68fa6d491e209131cea38c51 |
| ConSE(10) | 2.9 | Not recorded | 68fa62861e209131ce9f9e6e |
| Ourso-vs-o | 2.9 | Not recorded | 68fa62861e209131ce9f9e6e |
| Oursstruct | 2.9 | Not recorded | 68fa62861e209131ce9f9e6e |
| DGP | 2.9 | Not recorded | 68fa6d491e209131cea38c51 |
| ConSE(1) | 2.8 | Not recorded | 68fa62861e209131ce9f9e6e |
| ConSE(1000) | 2.8 | Not recorded | 68fa62861e209131ce9f9e6e |
| ConSE(10) \[33\] | 2.7 | Not recorded | 68fa62861e209131ce9f9e6e |
| ConSE(CV) | 2.6 | Not recorded | 68fa62861e209131ce9f9e6e |
| GCNZ | 2.2 | Not recorded | 68fa6d491e209131cea38c51 |

### DGP

Flat Hit@1: 6.3. Result ID: 698f9a363a6c314849fa7191.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65841e209131cea0abb5.md>)

Recorded evaluation configuration: {"Protocol":"ZSL"}. This record does not assert independent reproduction.

### ZSL-KG

Flat Hit@1: 6.3. Result ID: 698f9a363a6c314849fa7192.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Implementation](<https://github.com/BatsResearch/zsl-kg>)

Recorded evaluation configuration: {"Protocol":"ZSL"}. This record does not assert independent reproduction.

### SGCN

Flat Hit@1: 6. Result ID: 698f9a363a6c314849fa7190.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65841e209131cea0abb5.md>)

Recorded evaluation configuration: {"Protocol":"ZSL"}. This record does not assert independent reproduction.

### GCNZ

Flat Hit@1: 4.1. Result ID: 698f9a363a6c314849fa718f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64461e209131cea03a4f.md>)

Recorded evaluation configuration: {"Protocol":"ZSL"}. This record does not assert independent reproduction.

### ZSL-KG

Flat Hit@1: 3.4. Result ID: 698f9a363a6c314849fa7196.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Implementation](<https://github.com/BatsResearch/zsl-kg>)

Recorded evaluation configuration: {"Protocol":"GZSL"}. This record does not assert independent reproduction.

### SGCN

Flat Hit@1: 3.2. Result ID: 698f9a363a6c314849fa7194.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65841e209131cea0abb5.md>)

Recorded evaluation configuration: {"Protocol":"GZSL"}. This record does not assert independent reproduction.

### ConSE(10)

Flat Hit@1: 2.9. Result ID: 698bdaedecace47ef5592aa3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa632f1e209131ce9fd1fb.md>)

Recorded evaluation configuration: {"T":"10","Evaluation":"Zero-shot"}. This record does not assert independent reproduction.

### Ourso-vs-o

Flat Hit@1: 2.9. Result ID: 698bdaedecace47ef5592aa7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

Recorded evaluation configuration: {"Variant":"o-vs-o","Evaluation":"Zero-shot"}. This record does not assert independent reproduction.

### Oursstruct

Flat Hit@1: 2.9. Result ID: 698bdaedecace47ef5592aa8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

Recorded evaluation configuration: {"Variant":"struct","Evaluation":"Zero-shot"}. This record does not assert independent reproduction.

### DGP

Flat Hit@1: 2.9. Result ID: 698f9a363a6c314849fa7195.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65841e209131cea0abb5.md>)

Recorded evaluation configuration: {"Protocol":"GZSL"}. This record does not assert independent reproduction.

### ConSE(1)

Flat Hit@1: 2.8. Result ID: 698bdaedecace47ef5592aa2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa632f1e209131ce9fd1fb.md>)

Recorded evaluation configuration: {"T":"1","Evaluation":"Zero-shot"}. This record does not assert independent reproduction.

### ConSE(1000)

Flat Hit@1: 2.8. Result ID: 698bdaedecace47ef5592aa5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa632f1e209131ce9fd1fb.md>)

Recorded evaluation configuration: {"T":"1000","Evaluation":"Zero-shot"}. This record does not assert independent reproduction.

### ConSE(10) \[33\]

Flat Hit@1: 2.7. Result ID: 698bdaedecace47ef5592aa4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa632f1e209131ce9fd1fb.md>)

Recorded evaluation configuration: {"T":"10","Evaluation":"Zero-shot","Source":"Published Results"}. This record does not assert independent reproduction.

### ConSE(CV)

Flat Hit@1: 2.6. Result ID: 698bdaedecace47ef5592aa6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62861e209131ce9f9e6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa632f1e209131ce9fd1fb.md>)

Recorded evaluation configuration: {"T":"Class-wise CV","Evaluation":"Zero-shot"}. This record does not assert independent reproduction.

### GCNZ

Flat Hit@1: 2.2. Result ID: 698f9a363a6c314849fa7193.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6d491e209131cea38c51.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64461e209131cea03a4f.md>)

Recorded evaluation configuration: {"Protocol":"GZSL"}. This record does not assert independent reproduction.

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