# Zero-shot Learning on CUB (test)

Source: [Zero-shot Learning on CUB (test)](<https://www.sota2.com/research/sota/zero-shot-learning-on-cub-test>)

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

Data updated: 2026-02-26T06:04:10.567Z

Evaluation of Zero-shot Learning performance, reporting per-class accuracies (seen/novel) and harmonic mean, on the test split of the Caltech-UCSD Birds (CUB) 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 | Zero-shot Learning |
| Dataset (including any stated split) | CUB (test) |
| Metric | Seen Class Accuracy (a-\>s) |
| 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

- [Seen Class Accuracy (a-\>s)](<https://www.sota2.com/research/sota/zero-shot-learning-on-cub-test.md?metric=Seen%20Class%20Accuracy%20(a-%3Es)>): Higher is better; unit: PERCENT.

- [Zero-shot Accuracy (n-\>n)](<https://www.sota2.com/research/sota/zero-shot-learning-on-cub-test.md?metric=Zero-shot%20Accuracy%20(n-%3En)>): Higher is better; unit: PERCENT.

- [Novel Class Accuracy (a-\>n)](<https://www.sota2.com/research/sota/zero-shot-learning-on-cub-test.md?metric=Novel%20Class%20Accuracy%20(a-%3En)>): Higher is better; unit: PERCENT.

- [Harmonic Mean (H)](<https://www.sota2.com/research/sota/zero-shot-learning-on-cub-test.md?metric=Harmonic%20Mean%20(H)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Seen Class Accuracy (a-\>s) | Result date | Report paper |
| --- | --- | --- | --- |
| LFGAA+SA | 79.6 | Not recorded | 68fa6a061e209131cea258c7 |
| f-VAEGAN-d2 | 75.6 | Not recorded | 68fa6a061e209131cea258c7 |
| MLSE | 71.6 | Not recorded | 68fa6a061e209131cea258c7 |
| SMA | 71.3 | Not recorded | 68fa6a061e209131cea258c7 |
| APN | 69.3 | Not recorded | 68fa6a061e209131cea258c7 |
| AREN+CS | 69 | Not recorded | 68fa6a061e209131cea258c7 |
| f-VAEGAN-d2 | 63.8 | Not recorded | 68fa6a061e209131cea258c7 |
| Composer | 61.6 | Not recorded | 68fa6a061e209131cea258c7 |
| Self-Calibration | 59.6 | Not recorded | 68fa6a061e209131cea258c7 |
| Composer | 56.4 | Not recorded | 68fa6a061e209131cea258c7 |
| Self-Calibration | 55.4 | Not recorded | 68fa6a061e209131cea258c7 |
| f-Translator | 54.8 | Not recorded | 68fa6a061e209131cea258c7 |
| CADA-VAE | 53.5 | Not recorded | 68fa6a061e209131cea258c7 |
| Triplet Loss | 52.3 | Not recorded | 68fa6a061e209131cea258c7 |
| CVC | 47.6 | Not recorded | 68fa6a061e209131cea258c7 |

### LFGAA+SA

Seen Class Accuracy (a-\>s): 79.6. Result ID: 698dff9affc3901de09a09d2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Fine-tuned"}. This record does not assert independent reproduction.

### f-VAEGAN-d2

Seen Class Accuracy (a-\>s): 75.6. Result ID: 698dff9affc3901de09a09d4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65821e209131cea0aacd.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Fine-tuned","extra\_supervision":"human captions"}. This record does not assert independent reproduction.

### MLSE

Seen Class Accuracy (a-\>s): 71.6. Result ID: 698dff9affc3901de09a09c9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Pre-trained"}. This record does not assert independent reproduction.

### SMA

Seen Class Accuracy (a-\>s): 71.3. Result ID: 698dff9affc3901de09a09d1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa645f1e209131cea0417c.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Fine-tuned"}. This record does not assert independent reproduction.

### APN

Seen Class Accuracy (a-\>s): 69.3. Result ID: 698dff9affc3901de09a09d3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69eb1e209131cea24ee0.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Fine-tuned"}. This record does not assert independent reproduction.

### AREN+CS

Seen Class Accuracy (a-\>s): 69. Result ID: 698dff9affc3901de09a09d5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Fine-tuned"}. This record does not assert independent reproduction.

### f-VAEGAN-d2

Seen Class Accuracy (a-\>s): 63.8. Result ID: 698dff9affc3901de09a09cc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65821e209131cea0aacd.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Pre-trained","extra\_supervision":"human captions"}. This record does not assert independent reproduction.

### Composer

Seen Class Accuracy (a-\>s): 61.6. Result ID: 698dff9affc3901de09a09d7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Fine-tuned"}. This record does not assert independent reproduction.

### Self-Calibration

Seen Class Accuracy (a-\>s): 59.6. Result ID: 698dff9affc3901de09a09cf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Pre-trained"}. This record does not assert independent reproduction.

### Composer

Seen Class Accuracy (a-\>s): 56.4. Result ID: 698dff9affc3901de09a09d0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Pre-trained"}. This record does not assert independent reproduction.

### Self-Calibration

Seen Class Accuracy (a-\>s): 55.4. Result ID: 698dff9affc3901de09a09d6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Fine-tuned"}. This record does not assert independent reproduction.

### f-Translator

Seen Class Accuracy (a-\>s): 54.8. Result ID: 698dff9affc3901de09a09ce.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66af1e209131cea11bf3.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Pre-trained"}. This record does not assert independent reproduction.

### CADA-VAE

Seen Class Accuracy (a-\>s): 53.5. Result ID: 698dff9affc3901de09a09cd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa658a1e209131cea0af70.md>)

[Implementation](<https://github.com/edgarschnfld/CADA-VAE-PyTorch>)

Recorded evaluation configuration: {"Evaluation Setting":"Pre-trained"}. This record does not assert independent reproduction.

### Triplet Loss

Seen Class Accuracy (a-\>s): 52.3. Result ID: 698dff9affc3901de09a09cb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Pre-trained"}. This record does not assert independent reproduction.

### CVC

Seen Class Accuracy (a-\>s): 47.6. Result ID: 698dff9affc3901de09a09ca.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a061e209131cea258c7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66b81e209131cea12194.md>)

Recorded evaluation configuration: {"Evaluation Setting":"Pre-trained"}. This record does not assert independent reproduction.

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