Red-Teaming on HarmBench
96.3ASRAutoDAN-Turbo
Evaluation Results
| Method | Links | |
|---|---|---|
| AutoDAN-TurboTarget Model=Vicuna 7b2026.04 | 96.3 | |
| AutoDAN-TurboTarget Model=OpenChat 3.5 12102026.04 | 96.3 | |
| AutoDAN-TurboTarget Model=zephyr2026.04 | 96.3 | |
| STAR-TeamingTarget Model=zephyr2026.04 | 95.8 | |
| AutoDAN-TurboTarget Model=Solar 10.7B-Instruct2026.04 | 95.7 | |
| STAR-TeamingTarget Model=Vicuna 7b2026.04 | 93.8 | |
| STAR-TeamingTarget Model=Solar 10.7B-Instruct2026.04 | 93.8 | |
| STAR-TeamingTarget Model=OpenChat 3.5 12102026.04 | 93.5 | |
| STAR-TeamingTarget Model=Baichuan 2 7b2026.04 | 91.5 | |
| STAR-TeamingTarget Model=Qwen 7b chat2026.04 | 90.8 | |
| AutoDANTarget Model=Gemma3-4b-it2026.04 | 87.8 | |
| AutoDAN-TurboTarget Model=Baichuan 2 7b2026.04 | 83.8 | |
| AutoDAN-TurboTarget Model=GPT-4 Turbo 11062026.04 | 83.8 | |
| AutoDAN-TurboTarget Model=Qwen 7b chat2026.04 | 82.7 | |
| STAR-TeamingTarget Model=GPT-4 Turbo 11062026.04 | 78.1 | |
| STAR-TeamingTarget Model=GPT-4o2026.04 | 76.1 | |
| AutoDAN-TurboTarget Model=GPT-4o2026.04 | 76 | |
| AutoDANTarget Model=zephyr2026.04 | 75 | |
| STAR-TeamingTarget Model=Average2026.04 | 74.5 | |
| AutoDANTarget Model=OpenChat 3.5 12102026.04 | 73.5 | |
| AutoDANTarget Model=Solar 10.7B-Instruct2026.04 | 72.5 | |
| STAR-TeamingTarget Model=Qwen3-4b2026.04 | 72.5 | |
| STAR-TeamingTarget Model=Gemini Pro2026.04 | 72.5 | |
| STAR-TeamingTarget Model=Llama-2 13b chat2026.04 | 71.5 | |
| STAR-TeamingTarget Model=Qwen3-8b2026.04 | 71.5 | |
| STAR-TeamingTarget Model=Llama-2 7b chat2026.04 | 71 | |
| GCGTarget Model=zephyr2026.04 | 69.5 | |
| TAP-TTarget Model=zephyr2026.04 | 69.3 | |
| TAPTarget Model=Solar 10.7B-Instruct2026.04 | 66.5 | |
| TAPTarget Model=zephyr2026.04 | 66.5 | |
| GCGTarget Model=OpenChat 3.5 12102026.04 | 66.3 | |
| AutoDAN-TurboTarget Model=Gemini Pro2026.04 | 66.3 | |
| TAP-TTarget Model=OpenChat 3.5 12102026.04 | 66.1 | |
| AutoDANTarget Model=Vicuna 7b2026.04 | 66 | |
| HumanTarget Model=zephyr2026.04 | 66 | |
| TAPTarget Model=GPT-4o2026.04 | 66 | |
| TAP-TTarget Model=Solar 10.7B-Instruct2026.04 | 65.8 | |
| DirectTarget Model=zephyr2026.04 | 65.8 | |
| GCGTarget Model=Vicuna 7b2026.04 | 65.5 | |
| STAR-TeamingTarget Model=Gemma3-4b-it2026.04 | 64.7 | |
| TAPTarget Model=OpenChat 3.5 12102026.04 | 63.5 | |
| GBDATarget Model=zephyr2026.04 | 62.8 | |
| GCG-MTarget Model=zephyr2026.04 | 62.5 | |
| PEZTarget Model=zephyr2026.04 | 62.5 | |
| UATTarget Model=zephyr2026.04 | 62.3 | |
| SFSTarget Model=zephyr2026.04 | 62 | |
| GCG-MTarget Model=Solar 10.7B-Instruct2026.04 | 61.6 | |
| GCG-MTarget Model=Vicuna 7b2026.04 | 61.5 | |
| GCGTarget Model=Baichuan 2 7b2026.04 | 61.5 | |
| DirectTarget Model=Solar 10.7B-Instruct2026.04 | 61.3 | |
| HumanTarget Model=Solar 10.7B-Instruct2026.04 | 61.2 | |
| TAPTarget Model=Gemma3-12b-it2026.04 | 61.2 | |
| STAR-TeamingTarget Model=Llama-2 70b chat2026.04 | 61 | |
| GCG-TTarget Model=zephyr2026.04 | 61 | |
| AutoDAN-TurboTarget Model=Average2026.04 | 61 | |
| GCG-TTarget Model=Vicuna 7b2026.04 | 60.8 | |
| APTarget Model=zephyr2026.04 | 60.5 | |
| TAPTarget Model=Gemma3-4b-it2026.04 | 60.4 | |
| ZSTarget Model=zephyr2026.04 | 60 | |
| TAP-TTarget Model=Vicuna 7b2026.04 | 59.8 | |
| GCGTarget Model=Qwen 7b chat2026.04 | 59.2 | |
| PAIRTarget Model=Gemma3-4b-it2026.04 | 59.1 | |
| TAP-TTarget Model=Qwen 7b chat2026.04 | 59 | |
| GCG-TTarget Model=Solar 10.7B-Instruct2026.04 | 58.9 | |
| PAIRTarget Model=zephyr2026.04 | 58.8 | |
| TAP-TTarget Model=Baichuan 2 7b2026.04 | 58.5 | |
| TAP-TTarget Model=GPT-4 Turbo 11062026.04 | 58.5 | |
| SFSTarget Model=Solar 10.7B-Instruct2026.04 | 58.3 | |
| GCGTarget Model=Solar 10.7B-Instruct2026.04 | 57.5 | |
| GCG-TTarget Model=OpenChat 3.5 12102026.04 | 57.3 | |
| APTarget Model=OpenChat 3.5 12102026.04 | 57 | |
| PAIRTarget Model=Solar 10.7B-Instruct2026.04 | 56.8 | |
| STAR-TeamingTarget Model=Gemma3-12b-it2026.04 | 56.6 | |
| APTarget Model=Vicuna 7b2026.04 | 56.3 | |
| PEZTarget Model=Solar 10.7B-Instruct2026.04 | 56.1 | |
| ZSTarget Model=Solar 10.7B-Instruct2026.04 | 54.9 | |
| GCG-MTarget Model=OpenChat 3.5 12102026.04 | 54.6 | |
| GBDATarget Model=Solar 10.7B-Instruct2026.04 | 54.5 | |
| TAPTarget Model=Qwen3-8b2026.04 | 54.5 | |
| APTarget Model=Solar 10.7B-Instruct2026.04 | 54.3 | |
| UATTarget Model=Solar 10.7B-Instruct2026.04 | 54 | |
| PAIRTarget Model=Vicuna 7b2026.04 | 53.5 | |
| AutoDANTarget Model=Baichuan 2 7b2026.04 | 53.3 | |
| TAPTarget Model=Qwen 7b chat2026.04 | 53 | |
| PAIRTarget Model=GPT-4o2026.04 | 53 | |
| GCG-MTarget Model=Qwen 7b chat2026.04 | 52.5 | |
| SFSTarget Model=OpenChat 3.5 12102026.04 | 52.5 | |
| PAIRTarget Model=OpenChat 3.5 12102026.04 | 52.5 | |
| TAPTarget Model=Qwen3-4b2026.04 | 51.8 | |
| HumanTarget Model=OpenChat 3.5 12102026.04 | 51.3 | |
| TAPTarget Model=Vicuna 7b2026.04 | 51 | |
| TAPTarget Model=Baichuan 2 7b2026.04 | 51 | |
| PAIRTarget Model=Qwen 7b chat2026.04 | 50.2 | |
| APTarget Model=Qwen 7b chat2026.04 | 49.7 | |
| PAIRTarget Model=Gemma3-12b-it2026.04 | 48.5 | |
| APTarget Model=Baichuan 2 7b2026.04 | 48.3 | |
| AutoDAN-TurboTarget Model=Qwen3-4b2026.04 | 47.8 | |
| AutoDANTarget Model=Qwen 7b chat2026.04 | 47.3 | |
| AutoDANTarget Model=Gemma3-12b-it2026.04 | 47 | |
| GCG-TTarget Model=Baichuan 2 7b2026.04 | 46.4 |