# Text Adversarial Example Detection on Ag-News

Source: [Text Adversarial Example Detection on Ag-News](<https://www.sota2.com/research/sota/text-adversarial-example-detection-on-ag-news>)

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

Data updated: 2026-03-16T08:53:49.552Z

Evaluation of model performance in detecting adversarial text manipulations on the Ag-News dataset, measured using TPR@10, F1 score, and AUC.

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 | Text Adversarial Example Detection |
| Dataset (including any stated split) | Ag-News |
| Metric | TPR@10 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 28 of 28 |
| 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

- [TPR@10](<https://www.sota2.com/research/sota/text-adversarial-example-detection-on-ag-news.md?metric=TPR%4010>): Higher is better; unit: PERCENT.

- [F1 Score](<https://www.sota2.com/research/sota/text-adversarial-example-detection-on-ag-news.md?metric=F1%20Score>): Higher is better; unit: PERCENT.

- [AUC](<https://www.sota2.com/research/sota/text-adversarial-example-detection-on-ag-news.md?metric=AUC>): Higher is better; unit: PERCENT.

## Recorded results

| Method | TPR@10 | Result date | Report paper |
| --- | --- | --- | --- |
| RTD-Guard | 100 | Not recorded | 69b78e570636e5bf4afbd9fa |
| GradMask | 99 | Not recorded | 69b78e570636e5bf4afbd9fa |
| RTD-Guard | 98.1 | Not recorded | 69b78e570636e5bf4afbd9fa |
| RTD-Guard | 98.1 | Not recorded | 69b78e570636e5bf4afbd9fa |
| RTD-Guard | 97.5 | Not recorded | 69b78e570636e5bf4afbd9fa |
| WDR | 96.5 | Not recorded | 69b78e570636e5bf4afbd9fa |
| RDE | 95.8 | Not recorded | 69b78e570636e5bf4afbd9fa |
| GradMask | 95.4 | Not recorded | 69b78e570636e5bf4afbd9fa |
| WDR | 94.2 | Not recorded | 69b78e570636e5bf4afbd9fa |
| WDR | 94.1 | Not recorded | 69b78e570636e5bf4afbd9fa |
| RDE | 92.2 | Not recorded | 69b78e570636e5bf4afbd9fa |
| WDR | 91.4 | Not recorded | 69b78e570636e5bf4afbd9fa |
| FGWS | 91 | Not recorded | 69b78e570636e5bf4afbd9fa |
| GradMask | 90.9 | Not recorded | 69b78e570636e5bf4afbd9fa |
| GradMask | 90.6 | Not recorded | 69b78e570636e5bf4afbd9fa |
| RDE | 89.7 | Not recorded | 69b78e570636e5bf4afbd9fa |
| RDE | 89.6 | Not recorded | 69b78e570636e5bf4afbd9fa |
| FGWS | 82.4 | Not recorded | 69b78e570636e5bf4afbd9fa |
| MLE | 78.3 | Not recorded | 69b78e570636e5bf4afbd9fa |
| MLE | 73.5 | Not recorded | 69b78e570636e5bf4afbd9fa |
| MLE | 71.6 | Not recorded | 69b78e570636e5bf4afbd9fa |
| PPL | 69.5 | Not recorded | 69b78e570636e5bf4afbd9fa |
| MLE | 68.3 | Not recorded | 69b78e570636e5bf4afbd9fa |
| PPL | 66.2 | Not recorded | 69b78e570636e5bf4afbd9fa |
| FGWS | 65 | Not recorded | 69b78e570636e5bf4afbd9fa |
| FGWS | 64.1 | Not recorded | 69b78e570636e5bf4afbd9fa |
| PPL | 31.4 | Not recorded | 69b78e570636e5bf4afbd9fa |
| PPL | 23 | Not recorded | 69b78e570636e5bf4afbd9fa |

### RTD-Guard

TPR@10: 100. Result ID: 69b7c50647f34ce9249f89fa.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"TF-adj"}. This record does not assert independent reproduction.

### GradMask

TPR@10: 99. Result ID: 69b7c50647f34ce9249f89f6.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"TF-adj"}. This record does not assert independent reproduction.

### RTD-Guard

TPR@10: 98.1. Result ID: 69b7c50647f34ce9249f89f8.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"PWWS"}. This record does not assert independent reproduction.

### RTD-Guard

TPR@10: 98.1. Result ID: 69b7c50647f34ce9249f89f9.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"BAE"}. This record does not assert independent reproduction.

### RTD-Guard

TPR@10: 97.5. Result ID: 69b7c50647f34ce9249f89f7.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"TextFooler"}. This record does not assert independent reproduction.

### WDR

TPR@10: 96.5. Result ID: 69b7c50647f34ce9249f89ef.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fe91e209131cea481c0.md>)

[Implementation](<https://github.com/javirandor/wdr>)

Recorded evaluation configuration: {"Attack Method":"TextFooler"}. This record does not assert independent reproduction.

### RDE

TPR@10: 95.8. Result ID: 69b7c50647f34ce9249f89eb.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bf61e209131cea31401.md>)

Recorded evaluation configuration: {"Attack Method":"TextFooler"}. This record does not assert independent reproduction.

### GradMask

TPR@10: 95.4. Result ID: 69b7c50647f34ce9249f89f3.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"TextFooler"}. This record does not assert independent reproduction.

### WDR

TPR@10: 94.2. Result ID: 69b7c50647f34ce9249f89f0.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fe91e209131cea481c0.md>)

[Implementation](<https://github.com/javirandor/wdr>)

Recorded evaluation configuration: {"Attack Method":"PWWS"}. This record does not assert independent reproduction.

### WDR

TPR@10: 94.1. Result ID: 69b7c50647f34ce9249f89f2.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fe91e209131cea481c0.md>)

[Implementation](<https://github.com/javirandor/wdr>)

Recorded evaluation configuration: {"Attack Method":"TF-adj"}. This record does not assert independent reproduction.

### RDE

TPR@10: 92.2. Result ID: 69b7c50647f34ce9249f89ee.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bf61e209131cea31401.md>)

Recorded evaluation configuration: {"Attack Method":"TF-adj"}. This record does not assert independent reproduction.

### WDR

TPR@10: 91.4. Result ID: 69b7c50647f34ce9249f89f1.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6fe91e209131cea481c0.md>)

[Implementation](<https://github.com/javirandor/wdr>)

Recorded evaluation configuration: {"Attack Method":"BAE"}. This record does not assert independent reproduction.

### FGWS

TPR@10: 91. Result ID: 69b7c50647f34ce9249f89e8.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa692f1e209131cea20acf.md>)

Recorded evaluation configuration: {"Attack Method":"PWWS"}. This record does not assert independent reproduction.

### GradMask

TPR@10: 90.9. Result ID: 69b7c50647f34ce9249f89f4.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"PWWS"}. This record does not assert independent reproduction.

### GradMask

TPR@10: 90.6. Result ID: 69b7c50647f34ce9249f89f5.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"BAE"}. This record does not assert independent reproduction.

### RDE

TPR@10: 89.7. Result ID: 69b7c50647f34ce9249f89ec.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bf61e209131cea31401.md>)

Recorded evaluation configuration: {"Attack Method":"PWWS"}. This record does not assert independent reproduction.

### RDE

TPR@10: 89.6. Result ID: 69b7c50647f34ce9249f89ed.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6bf61e209131cea31401.md>)

Recorded evaluation configuration: {"Attack Method":"BAE"}. This record does not assert independent reproduction.

### FGWS

TPR@10: 82.4. Result ID: 69b7c50647f34ce9249f89e7.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa692f1e209131cea20acf.md>)

Recorded evaluation configuration: {"Attack Method":"TextFooler"}. This record does not assert independent reproduction.

### MLE

TPR@10: 78.3. Result ID: 69b7c50647f34ce9249f89e3.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64f91e209131cea077ac.md>)

[Implementation](<https://github.com/pokaxpoka/deep_Mahalanobis_detector>)

Recorded evaluation configuration: {"Attack Method":"TextFooler"}. This record does not assert independent reproduction.

### MLE

TPR@10: 73.5. Result ID: 69b7c50647f34ce9249f89e6.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64f91e209131cea077ac.md>)

[Implementation](<https://github.com/pokaxpoka/deep_Mahalanobis_detector>)

Recorded evaluation configuration: {"Attack Method":"TF-adj"}. This record does not assert independent reproduction.

### MLE

TPR@10: 71.6. Result ID: 69b7c50647f34ce9249f89e4.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64f91e209131cea077ac.md>)

[Implementation](<https://github.com/pokaxpoka/deep_Mahalanobis_detector>)

Recorded evaluation configuration: {"Attack Method":"PWWS"}. This record does not assert independent reproduction.

### PPL

TPR@10: 69.5. Result ID: 69b7c50647f34ce9249f89df.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"TextFooler"}. This record does not assert independent reproduction.

### MLE

TPR@10: 68.3. Result ID: 69b7c50647f34ce9249f89e5.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64f91e209131cea077ac.md>)

[Implementation](<https://github.com/pokaxpoka/deep_Mahalanobis_detector>)

Recorded evaluation configuration: {"Attack Method":"BAE"}. This record does not assert independent reproduction.

### PPL

TPR@10: 66.2. Result ID: 69b7c50647f34ce9249f89e0.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"PWWS"}. This record does not assert independent reproduction.

### FGWS

TPR@10: 65. Result ID: 69b7c50647f34ce9249f89e9.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa692f1e209131cea20acf.md>)

Recorded evaluation configuration: {"Attack Method":"BAE"}. This record does not assert independent reproduction.

### FGWS

TPR@10: 64.1. Result ID: 69b7c50647f34ce9249f89ea.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa692f1e209131cea20acf.md>)

Recorded evaluation configuration: {"Attack Method":"TF-adj"}. This record does not assert independent reproduction.

### PPL

TPR@10: 31.4. Result ID: 69b7c50647f34ce9249f89e2.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"TF-adj"}. This record does not assert independent reproduction.

### PPL

TPR@10: 23. Result ID: 69b7c50647f34ce9249f89e1.

[Reporting paper](<https://www.sota2.com/research/paper/69b78e570636e5bf4afbd9fa.md>)

Recorded evaluation configuration: {"Attack Method":"BAE"}. This record does not assert independent reproduction.

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