Jailbreaking on Attack Prompts In-domain
76ASR (TA)RogueMerge
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RogueMerge2026.06 | 76 | 92 | 54 | 87 | 77.3 | |
| LoBAM*2026.06 | 22 | 15 | 18 | 19 | 18.5 | |
| SFT2026.06 | 4 | 3 | 1 | 4 | 3 | |
| Clean2026.06 | 3 | 2 | 1 | 2 | 2 | |
| MergeHijacking*2026.06 | 3 | 5 | 2 | 5 | 3.8 |
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RogueMerge2026.06 | 76 | 92 | 54 | 87 | 77.3 | |
| LoBAM*2026.06 | 22 | 15 | 18 | 19 | 18.5 | |
| SFT2026.06 | 4 | 3 | 1 | 4 | 3 | |
| Clean2026.06 | 3 | 2 | 1 | 2 | 2 | |
| MergeHijacking*2026.06 | 3 | 5 | 2 | 5 | 3.8 |