# Text Classification on AG-News (Adversarial Robustness Metrics)

Source: [Text Classification on AG-News (Adversarial Robustness Metrics)](<https://www.sota2.com/research/sota/text-classification-on-ag-news-adversarial-robustness-metrics>)

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

Data updated: 2026-04-15T09:12:47.629Z

Evaluation of Text Classification performance on the AG-News dataset, focusing on adversarial robustness metrics such as Logit Flipping, SIA, and PGD-cw.

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 Classification |
| Dataset (including any stated split) | AG-News |
| Metric | Logit Flipping |
| 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

- [Logit Flipping](<https://www.sota2.com/research/sota/text-classification-on-ag-news-adversarial-robustness-metrics.md?metric=Logit%20Flipping>): Higher is better; unit: PERCENT.

- [SIA (BB)](<https://www.sota2.com/research/sota/text-classification-on-ag-news-adversarial-robustness-metrics.md?metric=SIA%20(BB)>): Higher is better; unit: PERCENT.

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

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

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

- [PGD (CW)](<https://www.sota2.com/research/sota/text-classification-on-ag-news-adversarial-robustness-metrics.md?metric=PGD%20(CW)>): Higher is better; unit: PERCENT.

- [Worst Case Robustness](<https://www.sota2.com/research/sota/text-classification-on-ag-news-adversarial-robustness-metrics.md?metric=Worst%20Case%20Robustness>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Logit Flipping | Result date | Report paper |
| --- | --- | --- | --- |
| DeepSet-TM | 85.7 | Not recorded | 68fa7c421e209131cea8d392 |
| Mean | 84.5 | Not recorded | 68fa7c421e209131cea8d392 |
| CWTM | 84.3 | Not recorded | 68fa7c421e209131cea8d392 |
| GM | 83 | Not recorded | 68fa7c421e209131cea8d392 |
| CWMed | 78.9 | Not recorded | 68fa7c421e209131cea8d392 |

### DeepSet-TM

Logit Flipping: 85.7. Result ID: 69df567f47f34ce924c84d72.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c421e209131cea8d392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c421e209131cea8d392.md>)

Recorded evaluation configuration: {"Heterogeneity (α)":"0.5","Clients (n)":"17","Adversaries (f)":"4"}. This record does not assert independent reproduction.

### Mean

Logit Flipping: 84.5. Result ID: 69df567f47f34ce924c84d6e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c421e209131cea8d392.md>)

Recorded evaluation configuration: {"Heterogeneity (α)":"0.5","Clients (n)":"17","Adversaries (f)":"4"}. This record does not assert independent reproduction.

### CWTM

Logit Flipping: 84.3. Result ID: 69df567f47f34ce924c84d71.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c421e209131cea8d392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69531e209131cea218fb.md>)

Recorded evaluation configuration: {"Heterogeneity (α)":"0.5","Clients (n)":"17","Adversaries (f)":"4"}. This record does not assert independent reproduction.

### GM

Logit Flipping: 83. Result ID: 69df567f47f34ce924c84d70.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c421e209131cea8d392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa703e1e209131cea49e91.md>)

Recorded evaluation configuration: {"Heterogeneity (α)":"0.5","Clients (n)":"17","Adversaries (f)":"4"}. This record does not assert independent reproduction.

### CWMed

Logit Flipping: 78.9. Result ID: 69df567f47f34ce924c84d6f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c421e209131cea8d392.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69531e209131cea218fb.md>)

Recorded evaluation configuration: {"Heterogeneity (α)":"0.5","Clients (n)":"17","Adversaries (f)":"4"}. This record does not assert independent reproduction.

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