# Jailbreak Attack Defense on SorryBench

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

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

Data updated: 2026-05-07T08:08:34.823Z

Evaluation of model defense capabilities against jailbreaking attacks using the SorryBench dataset, measured by FFR (Reasoning) and FFR (Final Answer).

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 Defense |
| Dataset (including any stated split) | SorryBench |
| Metric | FFR (Reasoning) |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 17 of 17 |
| 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

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

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

## Recorded results

| Method | FFR (Reasoning) | Result date | Report paper |
| --- | --- | --- | --- |
| Paraphrase | 56.8 | Not recorded | 68fa7a721e209131cea8346b |
| SmoothLLM | 54.5 | Not recorded | 68fa7a721e209131cea8346b |
| Paraphrase | 52.3 | Not recorded | 68fa7a721e209131cea8346b |
| No Defense | 50 | Not recorded | 68fa7a721e209131cea8346b |
| SmoothLLM | 50 | Not recorded | 68fa7a721e209131cea8346b |
| No Defense | 43.2 | Not recorded | 68fa7a721e209131cea8346b |
| Self-Reminder | 25 | Not recorded | 68fa7a721e209131cea8346b |
| SAFEPATH-ZS | 22.7 | Not recorded | 68fa7a721e209131cea8346b |
| SAFEPATH-ZS | 20.5 | Not recorded | 68fa7a721e209131cea8346b |
| Self-Reminder | 20.5 | Not recorded | 68fa7a721e209131cea8346b |
| ThinkingI | 15.9 | Not recorded | 68fa7a721e209131cea8346b |
| ReasoningGuard | 13.6 | Not recorded | 68fa7a721e209131cea8346b |
| RealSafe-R1 | 9.1 | Not recorded | 68fa7a721e209131cea8346b |
| ReasoningGuard | 9.1 | Not recorded | 68fa7a721e209131cea8346b |
| ThinkingI | 6.8 | Not recorded | 68fa7a721e209131cea8346b |
| SafeKey | 4.5 | Not recorded | 68fa7a721e209131cea8346b |
| RealSafe-R1 | 2.3 | Not recorded | 68fa7a721e209131cea8346b |

### Paraphrase

FFR (Reasoning): 56.8. Result ID: 69fc487847f34ce924ded92c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa707f1e209131cea4b7c6.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### SmoothLLM

FFR (Reasoning): 54.5. Result ID: 69fc487847f34ce924ded92e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

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

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### Paraphrase

FFR (Reasoning): 52.3. Result ID: 69fc487847f34ce924ded934.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa707f1e209131cea4b7c6.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-32B"}. This record does not assert independent reproduction.

### No Defense

FFR (Reasoning): 50. Result ID: 69fc487847f34ce924ded929.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### SmoothLLM

FFR (Reasoning): 50. Result ID: 69fc487847f34ce924ded936.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

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

Recorded evaluation configuration: {"Model":"R1-Qwen-32B"}. This record does not assert independent reproduction.

### No Defense

FFR (Reasoning): 43.2. Result ID: 69fc487847f34ce924ded932.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-32B"}. This record does not assert independent reproduction.

### Self-Reminder

FFR (Reasoning): 25. Result ID: 69fc487847f34ce924ded92d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### SAFEPATH-ZS

FFR (Reasoning): 22.7. Result ID: 69fc487847f34ce924ded938.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa790c1e209131cea7bde6.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-32B"}. This record does not assert independent reproduction.

### SAFEPATH-ZS

FFR (Reasoning): 20.5. Result ID: 69fc487847f34ce924ded930.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa790c1e209131cea7bde6.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### Self-Reminder

FFR (Reasoning): 20.5. Result ID: 69fc487847f34ce924ded935.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-32B"}. This record does not assert independent reproduction.

### ThinkingI

FFR (Reasoning): 15.9. Result ID: 69fc487847f34ce924ded92f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78b31e209131cea79d3c.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### ReasoningGuard

FFR (Reasoning): 13.6. Result ID: 69fc487847f34ce924ded931.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### RealSafe-R1

FFR (Reasoning): 9.1. Result ID: 69fc487847f34ce924ded92a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa780e1e209131cea76907.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### ReasoningGuard

FFR (Reasoning): 9.1. Result ID: 69fc487847f34ce924ded939.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-32B"}. This record does not assert independent reproduction.

### ThinkingI

FFR (Reasoning): 6.8. Result ID: 69fc487847f34ce924ded937.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78b31e209131cea79d3c.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-32B"}. This record does not assert independent reproduction.

### SafeKey

FFR (Reasoning): 4.5. Result ID: 69fc487847f34ce924ded92b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78be1e209131cea7a311.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-14B"}. This record does not assert independent reproduction.

### RealSafe-R1

FFR (Reasoning): 2.3. Result ID: 69fc487847f34ce924ded933.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a721e209131cea8346b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa780e1e209131cea76907.md>)

Recorded evaluation configuration: {"Model":"R1-Qwen-32B"}. This record does not assert independent reproduction.

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