# Anomaly Understanding on Anomaly-56K (test)

Source: [Anomaly Understanding on Anomaly-56K (test)](<https://www.sota2.com/research/sota/anomaly-understanding-on-anomaly-56k-test>)

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

Data updated: 2026-04-15T08:52:33.712Z

Evaluation of Anomaly Understanding performance on the test split of the Anomaly-56K dataset, measured using Localization Accuracy and ROUGE-L.

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 | Anomaly Understanding |
| Dataset (including any stated split) | Anomaly-56K (test) |
| Metric | Localization Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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

- [Localization Accuracy](<https://www.sota2.com/research/sota/anomaly-understanding-on-anomaly-56k-test.md?metric=Localization%20Accuracy>): Higher is better; unit: PERCENT.

- [ROUGE-L](<https://www.sota2.com/research/sota/anomaly-understanding-on-anomaly-56k-test.md?metric=ROUGE-L>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Localization Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| IAD-Unify (oracle) | 96.34 | Not recorded | 69df1b7a7d0172140ec7b83b |
| IAD-Unify (predicted) | 93.28 | Not recorded | 69df1b7a7d0172140ec7b83b |
| Qwen3.5-4B (no region input) | 73.83 | Not recorded | 69df1b7a7d0172140ec7b83b |
| Show-o2-7B | 44.83 | Not recorded | 69df1b7a7d0172140ec7b83b |
| Janus-Pro-7B | 40.2 | Not recorded | 69df1b7a7d0172140ec7b83b |

### IAD-Unify (oracle)

Localization Accuracy: 96.34. Result ID: 69df50d247f34ce924c82d12.

[Reporting paper](<https://www.sota2.com/research/paper/69df1b7a7d0172140ec7b83b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69df1b7a7d0172140ec7b83b.md>)

Recorded evaluation configuration: {"Region evidence source":"oracle","Backbone":"Qwen3.5-4B","Parameter count":"4B"}. This record does not assert independent reproduction.

### IAD-Unify (predicted)

Localization Accuracy: 93.28. Result ID: 69df50d247f34ce924c82d11.

[Reporting paper](<https://www.sota2.com/research/paper/69df1b7a7d0172140ec7b83b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69df1b7a7d0172140ec7b83b.md>)

Recorded evaluation configuration: {"Region evidence source":"predicted","Backbone":"Qwen3.5-4B","Parameter count":"4B"}. This record does not assert independent reproduction.

### Qwen3.5-4B (no region input)

Localization Accuracy: 73.83. Result ID: 69df50d247f34ce924c82d15.

[Reporting paper](<https://www.sota2.com/research/paper/69df1b7a7d0172140ec7b83b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa788e1e209131cea79116.md>)

Recorded evaluation configuration: {"Region evidence source":"none","Backbone":"Qwen3.5-4B","Parameter count":"4B"}. This record does not assert independent reproduction.

### Show-o2-7B

Localization Accuracy: 44.83. Result ID: 69df50d247f34ce924c82d13.

[Reporting paper](<https://www.sota2.com/research/paper/69df1b7a7d0172140ec7b83b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b1f1e209131cea86ae5.md>)

[Implementation](<https://github.com/showlab/Show-o>)

Recorded evaluation configuration: {"Backbone":"Show-o2-7B","Parameter count":"7B"}. This record does not assert independent reproduction.

### Janus-Pro-7B

Localization Accuracy: 40.2. Result ID: 69df50d247f34ce924c82d14.

[Reporting paper](<https://www.sota2.com/research/paper/69df1b7a7d0172140ec7b83b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa753c1e209131cea669f4.md>)

[Implementation](<https://github.com/deepseek-ai/Janus>)

Recorded evaluation configuration: {"Backbone":"Janus-Pro-7B","Parameter count":"7B"}. This record does not assert independent reproduction.

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