# Jailbreak Defense Performance on Jailbreak Attack Dataset

Source: [Jailbreak Defense Performance on Jailbreak Attack Dataset](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset>)

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

Data updated: 2026-02-26T05:40:31.907Z

Evaluation of model robustness and defense capabilities against jailbreak attacks using the Jailbreak Attack Dataset, measured by metrics such as DSR and BAR.

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 Defense Performance |
| Dataset (including any stated split) | Jailbreak Attack Dataset |
| Metric | DSR |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 33 of 33 |
| 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

- [DSR](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=DSR>): Higher is better; unit: PERCENT.

- [BAR](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=BAR>): Higher is better; unit: PERCENT.

- [ASR Reduction](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=ASR%20Reduction>): Higher is better; unit: PERCENT.

- [BAR Reduction](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=BAR%20Reduction>): Higher is better; unit: PERCENT.

- [MR](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=MR>): Higher is better; unit: PERCENT.

- [PE](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=PE>): Higher is better; unit: PERCENT.

- [AS](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=AS>): Higher is better; unit: PERCENT.

- [AG](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=AG>): Higher is better; unit: PERCENT.

- [Gradient Score](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=Gradient%20Score>): Higher is better; unit: PERCENT.

- [Perturbation Score](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=Perturbation%20Score>): Higher is better; unit: PERCENT.

- [Combination Score](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=Combination%20Score>): Higher is better; unit: SCALAR.

- [Average Performance](<https://www.sota2.com/research/sota/jailbreak-defense-performance-on-jailbreak-attack-dataset.md?metric=Average%20Performance>): Higher is better; unit: PERCENT.

## Recorded results

| Method | DSR | Result date | Report paper |
| --- | --- | --- | --- |
| Vanilla | 96.2 | Not recorded | 68fa73991e209131cea5d826 |
| Vanilla | 96.2 | Not recorded | 68fa73991e209131cea5d826 |
| Vanilla | 94.4 | Not recorded | 68fa73991e209131cea5d826 |
| Vanilla | 93.8 | Not recorded | 68fa73991e209131cea5d826 |
| CAVGAN | 91.12 | Not recorded | 68fa7a761e209131cea83657 |
| Self-Reminder | 80.6 | Not recorded | 68fa73991e209131cea5d826 |
| RA-LLM | 78.6 | Not recorded | 68fa7a761e209131cea83657 |
| CAVGAN | 77.22 | Not recorded | 68fa7a761e209131cea83657 |
| CAVGAN | 76.37 | Not recorded | 68fa7a761e209131cea83657 |
| Goal Prioritization | 75.6 | Not recorded | 68fa73991e209131cea5d826 |
| Self-Reminder | 73.8 | Not recorded | 68fa73991e209131cea5d826 |
| RA-LLM | 73.78 | Not recorded | 68fa7a761e209131cea83657 |
| RA-LLM | 71.18 | Not recorded | 68fa7a761e209131cea83657 |
| Vanilla | 70.6 | Not recorded | 68fa73991e209131cea5d826 |
| Self-Reminder | 68.8 | Not recorded | 68fa73991e209131cea5d826 |
| Smooth-llm | 54.22 | Not recorded | 68fa7a761e209131cea83657 |
| Smooth-llm | 52.07 | Not recorded | 68fa7a761e209131cea83657 |
| Smooth-llm | 48.97 | Not recorded | 68fa7a761e209131cea83657 |
| Self-Reminder | 37.5 | Not recorded | 68fa73991e209131cea5d826 |
| Goal Prioritization | 36.9 | Not recorded | 68fa73991e209131cea5d826 |
| Goal Prioritization | 26.9 | Not recorded | 68fa73991e209131cea5d826 |
| Original | 25.12 | Not recorded | 68fa7a761e209131cea83657 |
| Original | 18.59 | Not recorded | 68fa7a761e209131cea83657 |
| Original | 11.34 | Not recorded | 68fa7a761e209131cea83657 |
| Vanilla | 11 | Not recorded | 68fa73991e209131cea5d826 |
| Vanilla | 4 | Not recorded | 68fa73991e209131cea5d826 |
| Goal Prioritization | 2.5 | Not recorded | 68fa73991e209131cea5d826 |
| Self-Reminder | 2.5 | Not recorded | 68fa73991e209131cea5d826 |
| Goal Prioritization | 1.9 | Not recorded | 68fa73991e209131cea5d826 |
| Goal Prioritization | 1.9 | Not recorded | 68fa73991e209131cea5d826 |
| Self-Reminder | 1.2 | Not recorded | 68fa73991e209131cea5d826 |
| Self-Reminder | 0.6 | Not recorded | 68fa73991e209131cea5d826 |
| Goal Prioritization | 0.6 | Not recorded | 68fa73991e209131cea5d826 |

### Vanilla

DSR: 96.2. Result ID: 6992fa943a6c31484912cebb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

Recorded evaluation configuration: {"Model":"Vicuna-13B"}. This record does not assert independent reproduction.

### Vanilla

DSR: 96.2. Result ID: 6992fa943a6c31484912cebe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

Recorded evaluation configuration: {"Model":"Vicuna-33B"}. This record does not assert independent reproduction.

### Vanilla

DSR: 94.4. Result ID: 6992fa943a6c31484912ceb8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

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

### Vanilla

DSR: 93.8. Result ID: 6992fa943a6c31484912ceb2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

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

### CAVGAN

DSR: 91.12. Result ID: 698b78117a967eddb76eda28.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Implementation](<https://github.com/NLPGM/CAVGAN.>)

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

### Self-Reminder

DSR: 80.6. Result ID: 6992fa943a6c31484912cebf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

Recorded evaluation configuration: {"Model":"Vicuna-33B"}. This record does not assert independent reproduction.

### RA-LLM

DSR: 78.6. Result ID: 698b78117a967eddb76eda27.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d816.md>)

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

### CAVGAN

DSR: 77.22. Result ID: 698b78117a967eddb76eda2c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Implementation](<https://github.com/NLPGM/CAVGAN.>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B"}. This record does not assert independent reproduction.

### CAVGAN

DSR: 76.37. Result ID: 698b78117a967eddb76eda30.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Implementation](<https://github.com/NLPGM/CAVGAN.>)

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

### Goal Prioritization

DSR: 75.6. Result ID: 6992fa943a6c31484912ceba.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Implementation](<https://github.com/thu-coai/JailbreakDefense_GoalPriority>)

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

### Self-Reminder

DSR: 73.8. Result ID: 6992fa943a6c31484912ceb9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

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

### RA-LLM

DSR: 73.78. Result ID: 698b78117a967eddb76eda2b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d816.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B"}. This record does not assert independent reproduction.

### RA-LLM

DSR: 71.18. Result ID: 698b78117a967eddb76eda2f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d816.md>)

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

### Vanilla

DSR: 70.6. Result ID: 6992fa943a6c31484912ceb5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

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

### Self-Reminder

DSR: 68.8. Result ID: 6992fa943a6c31484912cebc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

Recorded evaluation configuration: {"Model":"Vicuna-13B"}. This record does not assert independent reproduction.

### Smooth-llm

DSR: 54.22. Result ID: 698b78117a967eddb76eda26.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73961e209131cea5d658.md>)

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

### Smooth-llm

DSR: 52.07. Result ID: 698b78117a967eddb76eda2e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73961e209131cea5d658.md>)

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

### Smooth-llm

DSR: 48.97. Result ID: 698b78117a967eddb76eda2a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73961e209131cea5d658.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B"}. This record does not assert independent reproduction.

### Self-Reminder

DSR: 37.5. Result ID: 6992fa943a6c31484912ceb3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

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

### Goal Prioritization

DSR: 36.9. Result ID: 6992fa943a6c31484912cebd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Implementation](<https://github.com/thu-coai/JailbreakDefense_GoalPriority>)

Recorded evaluation configuration: {"Model":"Vicuna-13B"}. This record does not assert independent reproduction.

### Goal Prioritization

DSR: 26.9. Result ID: 6992fa943a6c31484912cec0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Implementation](<https://github.com/thu-coai/JailbreakDefense_GoalPriority>)

Recorded evaluation configuration: {"Model":"Vicuna-33B"}. This record does not assert independent reproduction.

### Original

DSR: 25.12. Result ID: 698b78117a967eddb76eda25.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

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

### Original

DSR: 18.59. Result ID: 698b78117a967eddb76eda2d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

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

### Original

DSR: 11.34. Result ID: 698b78117a967eddb76eda29.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea83657.md>)

Recorded evaluation configuration: {"Model":"Llama3.1-8B"}. This record does not assert independent reproduction.

### Vanilla

DSR: 11. Result ID: 6992fa943a6c31484912cec4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

Recorded evaluation configuration: {"Model":"Llama2-13B-Chat"}. This record does not assert independent reproduction.

### Vanilla

DSR: 4. Result ID: 6992fa943a6c31484912cec1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

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

### Goal Prioritization

DSR: 2.5. Result ID: 6992fa943a6c31484912ceb4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Implementation](<https://github.com/thu-coai/JailbreakDefense_GoalPriority>)

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

### Self-Reminder

DSR: 2.5. Result ID: 6992fa943a6c31484912ceb6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

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

### Goal Prioritization

DSR: 1.9. Result ID: 6992fa943a6c31484912ceb7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Implementation](<https://github.com/thu-coai/JailbreakDefense_GoalPriority>)

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

### Goal Prioritization

DSR: 1.9. Result ID: 6992fa943a6c31484912cec6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Implementation](<https://github.com/thu-coai/JailbreakDefense_GoalPriority>)

Recorded evaluation configuration: {"Model":"Llama2-13B-Chat"}. This record does not assert independent reproduction.

### Self-Reminder

DSR: 1.2. Result ID: 6992fa943a6c31484912cec5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

Recorded evaluation configuration: {"Model":"Llama2-13B-Chat"}. This record does not assert independent reproduction.

### Self-Reminder

DSR: 0.6. Result ID: 6992fa943a6c31484912cec2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

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

### Goal Prioritization

DSR: 0.6. Result ID: 6992fa943a6c31484912cec3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73991e209131cea5d826.md>)

[Implementation](<https://github.com/thu-coai/JailbreakDefense_GoalPriority>)

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

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