# Adversarial Attack on MMLU 16 subjects filtered subset

Source: [Adversarial Attack on MMLU 16 subjects filtered subset](<https://www.sota2.com/research/sota/adversarial-attack-on-mmlu-16-subjects-filtered-subset>)

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

Data updated: 2026-02-26T05:24:06.529Z

Evaluation of adversarial attack robustness on the 16-subject filtered subset of the MMLU benchmark, reporting metrics such as ASR@30, SCE, and SEE.

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 Attack |
| Dataset (including any stated split) | MMLU 16 subjects filtered subset |
| Metric | ASR@30 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| 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

- [ASR@30](<https://www.sota2.com/research/sota/adversarial-attack-on-mmlu-16-subjects-filtered-subset.md?metric=ASR%4030>): Higher is better; unit: PERCENT.

- [SCE](<https://www.sota2.com/research/sota/adversarial-attack-on-mmlu-16-subjects-filtered-subset.md?metric=SCE>): Lower is better; unit: SCALAR.

- [SEE](<https://www.sota2.com/research/sota/adversarial-attack-on-mmlu-16-subjects-filtered-subset.md?metric=SEE>): Lower is better; unit: PERCENT.

## Recorded results

| Method | ASR@30 | Result date | Report paper |
| --- | --- | --- | --- |
| SECA | 81.24 | Not recorded | 68fa7be01e209131cea8af6e |
| SECA | 80.29 | Not recorded | 68fa7be01e209131cea8af6e |
| Raw | 63.52 | Not recorded | 68fa7be01e209131cea8af6e |
| Raw | 48.2 | Not recorded | 68fa7be01e209131cea8af6e |
| SECA | 36.86 | Not recorded | 68fa7be01e209131cea8af6e |
| Raw | 10.19 | Not recorded | 68fa7be01e209131cea8af6e |
| GCG | 9.86 | Not recorded | 68fa7be01e209131cea8af6e |
| GCG | 6.26 | Not recorded | 68fa7be01e209131cea8af6e |
| GCG | 0.57 | Not recorded | 68fa7be01e209131cea8af6e |

### SECA

ASR@30: 81.24. Result ID: 699a904d5698665ec41ae449.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

Recorded evaluation configuration: {"Target Model":"Llama-3-8B"}. This record does not assert independent reproduction.

### SECA

ASR@30: 80.29. Result ID: 699a904d5698665ec41ae446.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

Recorded evaluation configuration: {"Target Model":"Llama-3-3B"}. This record does not assert independent reproduction.

### Raw

ASR@30: 63.52. Result ID: 699a904d5698665ec41ae448.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690d1e209131cea1fcb0.md>)

[Implementation](<https://www.sota2.com/github.com/hendrycks/test>)

Recorded evaluation configuration: {"Target Model":"Llama-3-8B"}. This record does not assert independent reproduction.

### Raw

ASR@30: 48.2. Result ID: 699a904d5698665ec41ae445.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690d1e209131cea1fcb0.md>)

[Implementation](<https://www.sota2.com/github.com/hendrycks/test>)

Recorded evaluation configuration: {"Target Model":"Llama-3-3B"}. This record does not assert independent reproduction.

### SECA

ASR@30: 36.86. Result ID: 699a904d5698665ec41ae44c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

Recorded evaluation configuration: {"Target Model":"Qwen-2.5-7B"}. This record does not assert independent reproduction.

### Raw

ASR@30: 10.19. Result ID: 699a904d5698665ec41ae44b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa690d1e209131cea1fcb0.md>)

[Implementation](<https://www.sota2.com/github.com/hendrycks/test>)

Recorded evaluation configuration: {"Target Model":"Qwen-2.5-7B"}. This record does not assert independent reproduction.

### GCG

ASR@30: 9.86. Result ID: 699a904d5698665ec41ae44a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71861e209131cea51c3b.md>)

Recorded evaluation configuration: {"Target Model":"Llama-3-8B"}. This record does not assert independent reproduction.

### GCG

ASR@30: 6.26. Result ID: 699a904d5698665ec41ae447.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71861e209131cea51c3b.md>)

Recorded evaluation configuration: {"Target Model":"Llama-3-3B"}. This record does not assert independent reproduction.

### GCG

ASR@30: 0.57. Result ID: 699a904d5698665ec41ae44d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af6e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71861e209131cea51c3b.md>)

Recorded evaluation configuration: {"Target Model":"Qwen-2.5-7B"}. This record does not assert independent reproduction.

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