# Fine-grained Retrieval on CUB-200 (test)

Source: [Fine-grained Retrieval on CUB-200 (test)](<https://www.sota2.com/research/sota/fine-grained-retrieval-on-cub-200-test>)

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

Data updated: 2026-03-09T07:31:44.340Z

Evaluation of Fine-grained Retrieval performance on the test split of the CUB-200 dataset, measured using Recall at K metrics (R@1, R@5, R@10).

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 | Fine-grained retrieval |
| Dataset (including any stated split) | CUB-200 (test) |
| Metric | R@1 |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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

- [R@1](<https://www.sota2.com/research/sota/fine-grained-retrieval-on-cub-200-test.md?metric=R%401>): Higher is better; unit: SCALAR.

- [R@5](<https://www.sota2.com/research/sota/fine-grained-retrieval-on-cub-200-test.md?metric=R%405>): Higher is better; unit: SCALAR.

- [R@10](<https://www.sota2.com/research/sota/fine-grained-retrieval-on-cub-200-test.md?metric=R%4010>): Higher is better; unit: SCALAR.

## Recorded results

| Method | R@1 | Result date | Report paper |
| --- | --- | --- | --- |
| FAISS-HNSW | 0.768 | Not recorded | 69ae53fc2d603e6de3e42adc |
| Dense | 0.767 | Not recorded | 69ae53fc2d603e6de3e42adc |
| Two-stage BM25-V pipeline | 0.755 | Not recorded | 69ae53fc2d603e6de3e42adc |
| Two-stage BM25-V pipeline | 0.744 | Not recorded | 69ae53fc2d603e6de3e42adc |
| FAISS-IVF+PQ | 0.734 | Not recorded | 69ae53fc2d603e6de3e42adc |
| BM25-V | 0.472 | Not recorded | 69ae53fc2d603e6de3e42adc |

### FAISS-HNSW

R@1: 0.768. Result ID: 69ae77526dbb7075b9563b04.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64bf1e209131cea0645b.md>)

Recorded evaluation configuration: {"Pre-training":"SAE ImageNet-1K","Evaluation Protocol":"zero-shot","Interpretability":"None (Dense-only)"}. This record does not assert independent reproduction.

### Dense

R@1: 0.767. Result ID: 69ae77526dbb7075b9563b03.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

Recorded evaluation configuration: {"Pre-training":"SAE ImageNet-1K","Evaluation Protocol":"zero-shot","Interpretability":"None (Dense-only)"}. This record does not assert independent reproduction.

### Two-stage BM25-V pipeline

R@1: 0.755. Result ID: 69ae77526dbb7075b9563b08.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

Recorded evaluation configuration: {"Pre-training":"SAE ImageNet-1K","Evaluation Protocol":"zero-shot","Candidate pool size (K)":"200"}. This record does not assert independent reproduction.

### Two-stage BM25-V pipeline

R@1: 0.744. Result ID: 69ae77526dbb7075b9563b07.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

Recorded evaluation configuration: {"Pre-training":"SAE ImageNet-1K","Evaluation Protocol":"zero-shot","Candidate pool size (K)":"100"}. This record does not assert independent reproduction.

### FAISS-IVF+PQ

R@1: 0.734. Result ID: 69ae77526dbb7075b9563b05.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64671e209131cea046c9.md>)

Recorded evaluation configuration: {"Pre-training":"SAE ImageNet-1K","Evaluation Protocol":"zero-shot","Interpretability":"None (Dense-only)"}. This record does not assert independent reproduction.

### BM25-V

R@1: 0.472. Result ID: 69ae77526dbb7075b9563b06.

[Reporting paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ae53fc2d603e6de3e42adc.md>)

Recorded evaluation configuration: {"Pre-training":"SAE ImageNet-1K","Evaluation Protocol":"zero-shot"}. This record does not assert independent reproduction.

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