Jailbreak Defense Evaluation on EMRA MTA
1.1ASRCoopGuard
Evaluation Results
| Method | Links | |
|---|---|---|
| CoopGuardTarget LLM=GPT-52026.04 | 1.1 | |
| CoopGuardTarget LLM=Gemini-2.5-Pro2026.04 | 2.6 | |
| GoalPriorityTarget LLM=Gemini-2.5-Pro2026.04 | 2.8 | |
| GoalPriorityTarget LLM=GPT-52026.04 | 3 | |
| Self-ReminderTarget LLM=GPT-52026.04 | 4.8 | |
| Self-ReminderTarget LLM=Gemini-2.5-Pro2026.04 | 4.8 | |
| CoopGuardTarget LLM=DeepSeek-V32026.04 | 6 | |
| Self-ReminderTarget LLM=DeepSeek-V32026.04 | 6.8 | |
| PATTarget LLM=GPT-52026.04 | 7 | |
| GoalPriorityTarget LLM=DeepSeek-V32026.04 | 7 | |
| RPOTarget LLM=GPT-52026.04 | 7.8 | |
| RPOTarget LLM=Gemini-2.5-Pro2026.04 | 13.3 | |
| PATTarget LLM=Gemini-2.5-Pro2026.04 | 15 | |
| SecurityLinguaTarget LLM=GPT-52026.04 | 15.2 | |
| PATTarget LLM=DeepSeek-V32026.04 | 15.4 | |
| SecurityLinguaTarget LLM=DeepSeek-V32026.04 | 16.1 | |
| RPOTarget LLM=DeepSeek-V32026.04 | 17.5 | |
| SecurityLinguaTarget LLM=Gemini-2.5-Pro2026.04 | 18.3 |