Jailbreak Success Rate on AdvBench (Raw/Rules/ML Metrics)
3.8Raw Jailbreak Success RateGPT-4o
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-4oModel=GPT-4o2025.11 | 3.8 | 0 | 0 | 0 | |
| DeepSeek-R1Model=DeepSeek-R12025.11 | 2.3 | 0 | 0 | 0 | |
| Qwen3-235BModel=Qwen3-235B2025.11 | 1.9 | 0 | 0 | 0 |