Jailbreak Defense on Average Aggregate of 6 Attacks
0Attack Success Rate (ASR)SelfGrader
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SelfGraderBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=1.33, Memory Overhead (MB)=374.252026.04 | 0 | 0 | |
| SelfDefendBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=2.02, Memory Overhead (MB)=13137.25, Mode=Intent2026.04 | 2.34 | 27.72 | |
| SelfDefendBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=0.64, Memory Overhead (MB)=13130.75, Mode=Direct2026.04 | 4.96 | 38.42 | |
| WildGuardBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=0.94, Memory Overhead (MB)=13995.25, Evaluation Stage=Post2026.04 | 5.04 | 82.33 | |
| WildGuardBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=1.00, Memory Overhead (MB)=13939.25, Evaluation Stage=Pre2026.04 | 5.32 | 38.85 | |
| Prompt GuardBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=8.48, Memory Overhead (MB)=1154.582026.04 | 7.98 | 57.33 | |
| GradientCuffBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=80.70, Memory Overhead (MB)=1672.502026.04 | 9.85 | 50.25 | |
| Llama GuardBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=0.36, Memory Overhead (MB)=13592.75, Evaluation Stage=Pre2026.04 | 10.47 | 70.7 | |
| Llama GuardBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=0.45, Memory Overhead (MB)=13739.13, Evaluation Stage=Post2026.04 | 10.71 | 94.09 | |
| Qwen2.5-7B-Instruct (No Defense)Backbone=Qwen2.5-7B-Instruct, Latency (Sec.)=5.292026.04 | 14.93 | — | |
| Perplexity FilterBackbone=Qwen2.5-7B-Instruct, Latency (Sec.)=0.10, Memory Overhead (MB)=13134.882026.04 | 14.93 | 93.67 |