# Multi-turn Jailbreak Success Rate on JailbreakBench

Source: [Multi-turn Jailbreak Success Rate on JailbreakBench](<https://www.sota2.com/research/sota/multi-turn-jailbreak-success-rate-on-jailbreakbench>)

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

Data updated: 2026-04-14T11:40:21.583Z

Evaluation of model robustness against multi-turn jailbreaking attempts on the JailbreakBench dataset, measured using specific Attack Success Rate (ASR) methodologies (e.g., Salami, Crescendo, CIA).

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 | Multi-turn Jailbreak |
| Dataset (including any stated split) | JailbreakBench |
| Metric | Salami ASR (5-shot) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 2 of 2 |
| 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

- [Salami ASR (5-shot)](<https://www.sota2.com/research/sota/multi-turn-jailbreak-success-rate-on-jailbreakbench.md?metric=Salami%20ASR%20(5-shot)>): Higher is better; unit: PERCENT.

- [A-Salami ASR](<https://www.sota2.com/research/sota/multi-turn-jailbreak-success-rate-on-jailbreakbench.md?metric=A-Salami%20ASR>): Higher is better; unit: PERCENT.

- [Crescendo ASR](<https://www.sota2.com/research/sota/multi-turn-jailbreak-success-rate-on-jailbreakbench.md?metric=Crescendo%20ASR>): Higher is better; unit: PERCENT.

- [CIA ASR](<https://www.sota2.com/research/sota/multi-turn-jailbreak-success-rate-on-jailbreakbench.md?metric=CIA%20ASR>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Salami ASR (5-shot) | Result date | Report paper |
| --- | --- | --- | --- |
| GPT-4o (Baseline) | 87 | Not recorded | 69ddca4b7d0172140ec79ccb |
| CQA | 48 | Not recorded | 69ddca4b7d0172140ec79ccb |

### GPT-4o (Baseline)

Salami ASR (5-shot): 87. Result ID: 69de279547f34ce924c6b6f2.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca4b7d0172140ec79ccb.md>)

Recorded evaluation configuration: {"Defense":"w/o CQA","Target Model":"GPT-4o","Judge Model":"GPT-4o"}. This record does not assert independent reproduction.

### CQA

Salami ASR (5-shot): 48. Result ID: 69de279547f34ce924c6b6f3.

[Reporting paper](<https://www.sota2.com/research/paper/69ddca4b7d0172140ec79ccb.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ddca4b7d0172140ec79ccb.md>)

Recorded evaluation configuration: {"Defense":"w/ CQA","Target Model":"GPT-4o","Judge Model":"GPT-4o"}. This record does not assert independent reproduction.

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