# Jailbreak Detection on GCG\_adapt jailbreak prompts (Llama2-7B) (test)

Source: [Jailbreak Detection on GCG\_adapt jailbreak prompts (Llama2-7B) (test)](<https://www.sota2.com/research/sota/jailbreak-detection-on-gcg_adapt-jailbreak-prompts-llama2-7b-test>)

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

Data updated: 2026-06-08T07:52:32.503Z

Evaluation of Jailbreak Detection performance using the GCG\_adapt prompts specifically tailored for Llama2-7B, measured by ASR, TPR, FPR, and eASR.

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 Detection |
| Dataset (including any stated split) | GCG\_adapt jailbreak prompts (Llama2-7B) (test) |
| Metric | ASR |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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](<https://www.sota2.com/research/sota/jailbreak-detection-on-gcg_adapt-jailbreak-prompts-llama2-7b-test.md?metric=ASR>): Higher is better; unit: PERCENT.

- [TPR](<https://www.sota2.com/research/sota/jailbreak-detection-on-gcg_adapt-jailbreak-prompts-llama2-7b-test.md?metric=TPR>): Higher is better; unit: PERCENT.

- [FPR](<https://www.sota2.com/research/sota/jailbreak-detection-on-gcg_adapt-jailbreak-prompts-llama2-7b-test.md?metric=FPR>): Lower is better; unit: PERCENT.

- [eASR](<https://www.sota2.com/research/sota/jailbreak-detection-on-gcg_adapt-jailbreak-prompts-llama2-7b-test.md?metric=eASR>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ASR | Result date | Report paper |
| --- | --- | --- | --- |
| HiddenDetect | 81 | Not recorded | 6a264ca990966ade919beb19 |
| MTK | 72 | Not recorded | 6a264ca990966ade919beb19 |
| HSF | 67 | Not recorded | 6a264ca990966ade919beb19 |
| SaP | 62 | Not recorded | 6a264ca990966ade919beb19 |
| GradSafe | 57 | Not recorded | 6a264ca990966ade919beb19 |

### HiddenDetect

ASR: 81. Result ID: 6a2674acfbc373592c230342.

[Reporting paper](<https://www.sota2.com/research/paper/6a264ca990966ade919beb19.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79951e209131cea7e65c.md>)

Recorded evaluation configuration: {"Adaptive loss coefficient (λ)":"0.1","Target Model":"Llama2-7B"}. This record does not assert independent reproduction.

### MTK

ASR: 72. Result ID: 6a2674acfbc373592c230346.

[Reporting paper](<https://www.sota2.com/research/paper/6a264ca990966ade919beb19.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a264ca990966ade919beb19.md>)

Recorded evaluation configuration: {"Adaptive loss coefficient (λ)":"0.1","Adaptive loss":"J3","Target Model":"Llama2-7B"}. This record does not assert independent reproduction.

### HSF

ASR: 67. Result ID: 6a2674acfbc373592c230344.

[Reporting paper](<https://www.sota2.com/research/paper/6a264ca990966ade919beb19.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa792f1e209131cea7c7ef.md>)

Recorded evaluation configuration: {"Adaptive loss coefficient (λ)":"0.1","Target Model":"Llama2-7B"}. This record does not assert independent reproduction.

### SaP

ASR: 62. Result ID: 6a2674acfbc373592c230343.

[Reporting paper](<https://www.sota2.com/research/paper/6a264ca990966ade919beb19.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78e81e209131cea7b244.md>)

Recorded evaluation configuration: {"Adaptive loss coefficient (λ)":"0.1","Target Model":"Llama2-7B"}. This record does not assert independent reproduction.

### GradSafe

ASR: 57. Result ID: 6a2674acfbc373592c230345.

[Reporting paper](<https://www.sota2.com/research/paper/6a264ca990966ade919beb19.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa736c1e209131cea5c676.md>)

[Implementation](<https://github.com/xyq7/GradSafe>)

Recorded evaluation configuration: {"Adaptive loss coefficient (λ)":"0.05","Target Model":"Llama2-7B"}. This record does not assert independent reproduction.

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