Thinking Collapse Analysis on JailbreakBench (JBB)
0Thinking Collapse RateGCG
Evaluation Results
| Method | Links | |
|---|---|---|
| GCGModel=Qwen3 1.7B2026.03 | 0 | |
| PAIRModel=Qwen3 1.7B2026.03 | 0 | |
| AutoDANModel=Qwen3 1.7B2026.03 | 0 | |
| AutoDAN-TurboModel=Qwen3 1.7B2026.03 | 0 | |
| GCGModel=Qwen3 4B2026.03 | 0 | |
| PAIRModel=Qwen3 4B2026.03 | 0 | |
| AutoDANModel=Qwen3 4B2026.03 | 0 | |
| AutoDAN-TurboModel=Qwen3 4B2026.03 | 0 | |
| GCGModel=Qwen3 8B2026.03 | 0 | |
| PAIRModel=Qwen3 8B2026.03 | 0 | |
| AutoDANModel=Qwen3 8B2026.03 | 0 | |
| AutoDAN-TurboModel=Qwen3 8B2026.03 | 0 | |
| AutoDANModel=DeepSeek2026.03 | 0 | |
| AutoDAN-TurboModel=DeepSeek2026.03 | 0 | |
| PAIRModel=Qwen3-Max2026.03 | 0 | |
| AutoDANModel=Qwen3-Max2026.03 | 0 | |
| AutoDAN-TurboModel=Qwen3-Max2026.03 | 0 | |
| PAIRModel=DeepSeek2026.03 | 0.4 | |
| GCGModel=DeepSeek2026.03 | 0.5 | |
| GCGModel=Qwen3-Max2026.03 | 0.5 | |
| MS_ReverseModel=Qwen3-Max2026.03 | 1 | |
| MS_ReverseModel=Qwen3 1.7B2026.03 | 8 | |
| MS_ReverseModel=DeepSeek2026.03 | 8 | |
| MS_ReverseModel=Qwen3 8B2026.03 | 9 | |
| MS_ReverseModel=Qwen3 4B2026.03 | 17 |