# Jailbreak Attack on Harmbench Malicious (full)

Source: [Jailbreak Attack on Harmbench Malicious (full)](<https://www.sota2.com/research/sota/jailbreak-attack-on-harmbench-malicious-full>)

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

Data updated: 2026-02-26T05:46:23.495Z

Evaluation of model robustness against jailbreak attacks using the full set of malicious prompts from the HarmBench dataset, measured by Attack Success Rate (ASR) and Harmful Score.

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 | Jailbreak Attack |
| Dataset (including any stated split) | Harmbench Malicious (full) |
| Metric | Harmful Score |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 14 of 14 |
| 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

- [Harmful Score](<https://www.sota2.com/research/sota/jailbreak-attack-on-harmbench-malicious-full.md?metric=Harmful%20Score>): Lower is better; unit: SCALAR.

- [ASR](<https://www.sota2.com/research/sota/jailbreak-attack-on-harmbench-malicious-full.md?metric=ASR>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Harmful Score | Result date | Report paper |
| --- | --- | --- | --- |
| Irrelevant Fine-tuning | 1.13 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Irrelevant Fine-tuning | 1.74 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Irrelevant Fine-tuning | 2.27 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Irrelevant Fine-tuning | 2.7 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Irrelevant Fine-tuning | 2.85 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Irrelevant Fine-tuning | 2.89 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Irrelevant Fine-tuning | 3.12 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Malicious Fine-tuning | 4.38 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Malicious Fine-tuning | 4.79 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Malicious Fine-tuning | 4.8 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Malicious Fine-tuning | 4.82 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Malicious Fine-tuning | 4.85 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Malicious Fine-tuning | 4.87 | Not recorded | 69705dafe25ad2fa9d1ed32e |
| Malicious Fine-tuning | 4.91 | Not recorded | 69705dafe25ad2fa9d1ed32e |

### Irrelevant Fine-tuning

Harmful Score: 1.13. Result ID: 6996eb8c5698665ec43810a3.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

Recorded evaluation configuration: {"Model":"Llama2-7b"}. This record does not assert independent reproduction.

### Irrelevant Fine-tuning

Harmful Score: 1.74. Result ID: 6996eb8c5698665ec43810a9.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

Recorded evaluation configuration: {"Model":"GPT-3.5-Turbo"}. This record does not assert independent reproduction.

### Irrelevant Fine-tuning

Harmful Score: 2.27. Result ID: 6996eb8c5698665ec43810a6.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7b"}. This record does not assert independent reproduction.

### Irrelevant Fine-tuning

Harmful Score: 2.7. Result ID: 6996eb8c5698665ec43810a7.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

Recorded evaluation configuration: {"Model":"Llama3-70b"}. This record does not assert independent reproduction.

### Irrelevant Fine-tuning

Harmful Score: 2.85. Result ID: 6996eb8c5698665ec43810a8.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

Recorded evaluation configuration: {"Model":"GPT-4o-mini"}. This record does not assert independent reproduction.

### Irrelevant Fine-tuning

Harmful Score: 2.89. Result ID: 6996eb8c5698665ec43810a5.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3b"}. This record does not assert independent reproduction.

### Irrelevant Fine-tuning

Harmful Score: 3.12. Result ID: 6996eb8c5698665ec43810a4.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

Recorded evaluation configuration: {"Model":"Llama3-8b"}. This record does not assert independent reproduction.

### Malicious Fine-tuning

Harmful Score: 4.38. Result ID: 6996eb8c5698665ec4381096.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e11e209131cea5ea4c.md>)

Recorded evaluation configuration: {"Model":"Llama3-8b"}. This record does not assert independent reproduction.

### Malicious Fine-tuning

Harmful Score: 4.79. Result ID: 6996eb8c5698665ec4381098.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e11e209131cea5ea4c.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7b"}. This record does not assert independent reproduction.

### Malicious Fine-tuning

Harmful Score: 4.8. Result ID: 6996eb8c5698665ec438109b.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e11e209131cea5ea4c.md>)

Recorded evaluation configuration: {"Model":"GPT-3.5-Turbo"}. This record does not assert independent reproduction.

### Malicious Fine-tuning

Harmful Score: 4.82. Result ID: 6996eb8c5698665ec4381095.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e11e209131cea5ea4c.md>)

Recorded evaluation configuration: {"Model":"Llama2-7b"}. This record does not assert independent reproduction.

### Malicious Fine-tuning

Harmful Score: 4.85. Result ID: 6996eb8c5698665ec438109a.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e11e209131cea5ea4c.md>)

Recorded evaluation configuration: {"Model":"GPT-4o-mini"}. This record does not assert independent reproduction.

### Malicious Fine-tuning

Harmful Score: 4.87. Result ID: 6996eb8c5698665ec4381097.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e11e209131cea5ea4c.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3b"}. This record does not assert independent reproduction.

### Malicious Fine-tuning

Harmful Score: 4.91. Result ID: 6996eb8c5698665ec4381099.

[Reporting paper](<https://www.sota2.com/research/paper/69705dafe25ad2fa9d1ed32e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73e11e209131cea5ea4c.md>)

Recorded evaluation configuration: {"Model":"Llama3-70b"}. This record does not assert independent reproduction.

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