Jailbreak Attack Defense on FORTRESS
9.8ASRReasoningGuard
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ReasoningGuardModel Scale=Qwen3-14B2026.01 | 9.8 | — | |
| Self-GuardModel Scale=Qwen3-14B2026.01 | 11.2 | — | |
| Self-ReminderModel Scale=Qwen3-14B2026.01 | 14.4 | — | |
| ReasoningGuardModel Scale=Qwen3-8B2026.01 | 14.8 | — | |
| Self-GuardModel Scale=Qwen3-8B2026.01 | 16 | — | |
| SafeKeyModel Scale=Qwen3-4B2026.01 | 16.8 | — | |
| ReasoningGuardModel Scale=Qwen3-4B2026.01 | 16.8 | — | |
| Self-GuardModel Scale=Qwen3-4B2026.01 | 17.4 | — | |
| Self-ReminderModel Scale=Qwen3-8B2026.01 | 19.2 | — | |
| SafeKeyModel Scale=Qwen3-14B2026.01 | 19.4 | — | |
| STAR-1Model Scale=Qwen3-14B2026.01 | 20.2 | — | |
| SafeKeyModel Scale=Qwen3-8B2026.01 | 20.8 | — | |
| STAR-1Model Scale=Qwen3-4B2026.01 | 24.4 | — | |
| Self-ReminderModel Scale=Qwen3-4B2026.01 | 24.4 | — | |
| STAR-1Model Scale=Qwen3-8B2026.01 | 25.6 | — | |
| SafeChainModel Scale=Qwen3-4B2026.01 | 39.8 | — | |
| SafeChainModel Scale=Qwen3-14B2026.01 | 41.2 | — | |
| Alpha-steerModel Scale=Qwen3-14B2026.01 | 41.4 | — | |
| SafeChainModel Scale=Qwen3-8B2026.01 | 44.6 | — | |
| Alpha-steerModel Scale=Qwen3-4B2026.01 | 45.2 | — | |
| InstructModel Scale=Qwen3-14B2026.01 | 47.2 | — | |
| InstructModel Scale=Qwen3-4B2026.01 | 52 | — | |
| InstructModel Scale=Qwen3-8B2026.01 | 52.2 | — | |
| Alpha-steerModel Scale=Qwen3-8B2026.01 | 56.9 | — |