# Pixel-level Image Classification on Image 4K Long Range Arena (LRA) (test)

Source: [Pixel-level Image Classification on Image 4K Long Range Arena (LRA) (test)](<https://www.sota2.com/research/sota/pixel-level-image-classification-on-image-4k-long-range-arena-lra-test>)

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

Data updated: 2026-02-26T05:33:56.792Z

Evaluation of pixel-level image classification performance using Accuracy (%) on the test split of the 4K configuration of the Long Range Arena (LRA) Image task.

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 | Pixel-level Image Classification |
| Dataset (including any stated split) | Image 4K Long Range Arena (LRA) (test) |
| Metric | Accuracy (%) |
| Unit | PERCENT |
| 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

- [Accuracy (%)](<https://www.sota2.com/research/sota/pixel-level-image-classification-on-image-4k-long-range-arena-lra-test.md?metric=Accuracy%20(%25)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy (%) | Result date | Report paper |
| --- | --- | --- | --- |
| Efficient Attention + GFSA | 40.4 | Not recorded | 68fa758e1e209131cea68985 |
| Transformer+GFSA | 40.2 | Not recorded | 68fa758e1e209131cea68985 |
| Efficient Attention | 40.2 | Not recorded | 68fa758e1e209131cea68985 |
| YOSO-E | 39.8 | Not recorded | 68fa758e1e209131cea68985 |
| Transformer | 38.2 | Not recorded | 68fa758e1e209131cea68985 |
| Linformer | 37.8 | Not recorded | 68fa758e1e209131cea68985 |

### Efficient Attention + GFSA

Accuracy (%): 40.4. Result ID: 6993c6c63a6c3148491871a6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758e1e209131cea68985.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758e1e209131cea68985.md>)

Recorded evaluation configuration: {"Runtime (s per 1,000 steps)":135.8,"GPU usage (GB)":1.33}. This record does not assert independent reproduction.

### Transformer+GFSA

Accuracy (%): 40.2. Result ID: 6993c6c63a6c3148491871a2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758e1e209131cea68985.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758e1e209131cea68985.md>)

Recorded evaluation configuration: {"Runtime (s per 1,000 steps)":737.2,"GPU usage (GB)":11.2}. This record does not assert independent reproduction.

### Efficient Attention

Accuracy (%): 40.2. Result ID: 6993c6c63a6c3148491871a5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758e1e209131cea68985.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71d71e209131cea5346b.md>)

[Implementation](<https://github.com/cmsflash/efficient-attention>)

Recorded evaluation configuration: {"Runtime (s per 1,000 steps)":121.1,"GPU usage (GB)":1.14}. This record does not assert independent reproduction.

### YOSO-E

Accuracy (%): 39.8. Result ID: 6993c6c63a6c3148491871a4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758e1e209131cea68985.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b421e209131cea2d02f.md>)

Recorded evaluation configuration: {"Runtime (s per 1,000 steps)":114.2,"GPU usage (GB)":1.42}. This record does not assert independent reproduction.

### Transformer

Accuracy (%): 38.2. Result ID: 6993c6c63a6c3148491871a1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758e1e209131cea68985.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa703d1e209131cea49db1.md>)

[Implementation](<https://github.com/tensorflow/tensor2tensor>)

Recorded evaluation configuration: {"Runtime (s per 1,000 steps)":345.1,"GPU usage (GB)":5.88}. This record does not assert independent reproduction.

### Linformer

Accuracy (%): 37.8. Result ID: 6993c6c63a6c3148491871a3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa758e1e209131cea68985.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67bb1e209131cea182c9.md>)

Recorded evaluation configuration: {"Runtime (s per 1,000 steps)":158.5,"GPU usage (GB)":3.45}. This record does not assert independent reproduction.

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