# Adversarial Detection on Wikipedia HiSPA word-shuffle benign adversarial (N=500 held-out)

Source: [Adversarial Detection on Wikipedia HiSPA word-shuffle benign adversarial (N=500 held-out)](<https://www.sota2.com/research/sota/adversarial-detection-on-wikipedia-hispa-word-shuffle-benign-adversarial-n-500-held-out>)

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

Data updated: 2026-03-16T08:49:03.920Z

Evaluation of models designed to detect benign adversarial examples generated via word-shuffling techniques on a held-out subset (N=500) of the Wikipedia HiSPA dataset. Metrics include F1 score and correlation shifts (rho).

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 | Adversarial Detection |
| Dataset (including any stated split) | Wikipedia HiSPA word-shuffle benign adversarial N=500 per model (held-out) |
| Metric | F1 Score |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 3 of 3 |
| 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

- [F1 Score](<https://www.sota2.com/research/sota/adversarial-detection-on-wikipedia-hispa-word-shuffle-benign-adversarial-n-500-held-out.md?metric=F1%20Score>): Higher is better; unit: PERCENT.

- [False Positive Rate (FPR)](<https://www.sota2.com/research/sota/adversarial-detection-on-wikipedia-hispa-word-shuffle-benign-adversarial-n-500-held-out.md?metric=False%20Positive%20Rate%20(FPR)>): Lower is better; unit: PERCENT.

- [Correlation (rho) Benign](<https://www.sota2.com/research/sota/adversarial-detection-on-wikipedia-hispa-word-shuffle-benign-adversarial-n-500-held-out.md?metric=Correlation%20(rho)%20Benign>): Higher is better; unit: SCALAR.

- [Correlation (rho) Adversarial](<https://www.sota2.com/research/sota/adversarial-detection-on-wikipedia-hispa-word-shuffle-benign-adversarial-n-500-held-out.md?metric=Correlation%20(rho)%20Adversarial>): Higher is better; unit: SCALAR.

- [Delta Correlation (rho)](<https://www.sota2.com/research/sota/adversarial-detection-on-wikipedia-hispa-word-shuffle-benign-adversarial-n-500-held-out.md?metric=Delta%20Correlation%20(rho)>): Lower is better; unit: SCALAR.

## Recorded results

| Method | F1 Score | Result date | Report paper |
| --- | --- | --- | --- |
| SpectralGuard | 65 | Not recorded | 69b78e970636e5bf4afbdc5f |
| SpectralGuard | 61.9 | Not recorded | 69b78e970636e5bf4afbdc5f |
| SpectralGuard | 59.1 | Not recorded | 69b78e970636e5bf4afbdc5f |

### SpectralGuard

F1 Score: 65. Result ID: 69b7c30247f34ce9249f781b.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e970636e5bf4afbdc5f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b78e970636e5bf4afbdc5f.md>)

Recorded evaluation configuration: {"Model":"Mamba-2.8B","Params":"2,897M"}. This record does not assert independent reproduction.

### SpectralGuard

F1 Score: 61.9. Result ID: 69b7c30247f34ce9249f7819.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e970636e5bf4afbdc5f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b78e970636e5bf4afbdc5f.md>)

Recorded evaluation configuration: {"Model":"Mamba-130M","Params":"130M"}. This record does not assert independent reproduction.

### SpectralGuard

F1 Score: 59.1. Result ID: 69b7c30247f34ce9249f781a.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e970636e5bf4afbdc5f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b78e970636e5bf4afbdc5f.md>)

Recorded evaluation configuration: {"Model":"Mamba-1.4B","Params":"1,475M"}. This record does not assert independent reproduction.

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