# Harmfulness Evaluation on AdvBench-audio Harmful

Source: [Harmfulness Evaluation on AdvBench-audio Harmful](<https://www.sota2.com/research/sota/harmfulness-evaluation-on-advbench-audio-harmful>)

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

Data updated: 2026-05-08T08:27:23.072Z

Evaluation of model performance on the AdvBench-audio Harmful dataset for Harmfulness Evaluation, measured using ASR.

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 | Harmfulness Evaluation |
| Dataset (including any stated split) | AdvBench-audio Harmful |
| Metric | ASR Score |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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 Score](<https://www.sota2.com/research/sota/harmfulness-evaluation-on-advbench-audio-harmful.md?metric=ASR%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | ASR Score | Result date | Report paper |
| --- | --- | --- | --- |
| MDSteer-h2s | 26.35 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| MDSteer-c2r | 23.46 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| MDSteer-h2s | 10.38 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| RRS | 3.27 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| No Defense | 2.88 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| No Defense | 2.88 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| AdaShield | 1.15 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| FSD | 0.78 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| SARSteer | 0.58 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| SARSteer | 0.58 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| No Defense | 0 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| AdaShield | 0 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| FSD | 0 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| MDSteer-c2r | 0 | Not recorded | 68fa7c5d1e209131cea8dc06 |
| SARSteer | 0 | Not recorded | 68fa7c5d1e209131cea8dc06 |

### MDSteer-h2s

ASR Score: 26.35. Result ID: 69fd9df447f34ce924e09041.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa757b1e209131cea68414.md>)

[Implementation](<https://github.com/andyrdt/refusal_direction>)

Recorded evaluation configuration: {"Model":"Qwen2-Audio"}. This record does not assert independent reproduction.

### MDSteer-c2r

ASR Score: 23.46. Result ID: 69fd9df447f34ce924e09042.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b641e209131cea88272.md>)

Recorded evaluation configuration: {"Model":"Qwen2-Audio"}. This record does not assert independent reproduction.

### MDSteer-h2s

ASR Score: 10.38. Result ID: 69fd9df447f34ce924e09047.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa757b1e209131cea68414.md>)

[Implementation](<https://github.com/andyrdt/refusal_direction>)

Recorded evaluation configuration: {"Model":"Kimi-Audio"}. This record does not assert independent reproduction.

### RRS

ASR Score: 3.27. Result ID: 69fd9df447f34ce924e0906c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78c11e209131cea7a528.md>)

Recorded evaluation configuration: {"Defense runtime (s)":"1145"}. This record does not assert independent reproduction.

### No Defense

ASR Score: 2.88. Result ID: 69fd9df447f34ce924e0903e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

Recorded evaluation configuration: {"Model":"Qwen2-Audio"}. This record does not assert independent reproduction.

### No Defense

ASR Score: 2.88. Result ID: 69fd9df447f34ce924e0906b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### AdaShield

ASR Score: 1.15. Result ID: 69fd9df447f34ce924e0903f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa727e1e209131cea56bc0.md>)

Recorded evaluation configuration: {"Model":"Qwen2-Audio"}. This record does not assert independent reproduction.

### FSD

ASR Score: 0.78. Result ID: 69fd9df447f34ce924e09040.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76b01e209131cea6e7cf.md>)

Recorded evaluation configuration: {"Model":"Qwen2-Audio"}. This record does not assert independent reproduction.

### SARSteer

ASR Score: 0.58. Result ID: 69fd9df447f34ce924e09043.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

Recorded evaluation configuration: {"Model":"Qwen2-Audio"}. This record does not assert independent reproduction.

### SARSteer

ASR Score: 0.58. Result ID: 69fd9df447f34ce924e0906d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

Recorded evaluation configuration: {"Defense runtime (s)":"266"}. This record does not assert independent reproduction.

### No Defense

ASR Score: 0. Result ID: 69fd9df447f34ce924e09044.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

Recorded evaluation configuration: {"Model":"Kimi-Audio"}. This record does not assert independent reproduction.

### AdaShield

ASR Score: 0. Result ID: 69fd9df447f34ce924e09045.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa727e1e209131cea56bc0.md>)

Recorded evaluation configuration: {"Model":"Kimi-Audio"}. This record does not assert independent reproduction.

### FSD

ASR Score: 0. Result ID: 69fd9df447f34ce924e09046.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76b01e209131cea6e7cf.md>)

Recorded evaluation configuration: {"Model":"Kimi-Audio"}. This record does not assert independent reproduction.

### MDSteer-c2r

ASR Score: 0. Result ID: 69fd9df447f34ce924e09048.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7b641e209131cea88272.md>)

Recorded evaluation configuration: {"Model":"Kimi-Audio"}. This record does not assert independent reproduction.

### SARSteer

ASR Score: 0. Result ID: 69fd9df447f34ce924e09049.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c5d1e209131cea8dc06.md>)

Recorded evaluation configuration: {"Model":"Kimi-Audio"}. This record does not assert independent reproduction.

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