Jailbreak Attack Resilience on HarmBench Misinfo. & Disinfo.
80ASRPhi-3 Mini
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Phi-3 MiniN=202026.05 | 80 | 1 | 25 | 20 | |
| Qwen 3 (4B)N=202026.05 | 15 | 8 | 95 | 85 | |
| Llama 3.2 (3B)N=202026.05 | 5 | 7 | 100 | 95 |
| Method | Links | ||||
|---|---|---|---|---|---|
| Phi-3 MiniN=202026.05 | 80 | 1 | 25 | 20 | |
| Qwen 3 (4B)N=202026.05 | 15 | 8 | 95 | 85 | |
| Llama 3.2 (3B)N=202026.05 | 5 | 7 | 100 | 95 |