Jailbreaking on SorryBench
43.6LG4 ASRCRaFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CRaFTTarget Model=google/gemma-3-1b-it, Decoding=greedy, max_new_tokens=5122026.04 | 43.6 | 2.77 | |
| Refusal-DirectionTarget Model=google/gemma-3-1b-it, Decoding=greedy, max_new_tokens=5122026.04 | 22 | 0.44 | |
| Refusal-SAETarget Model=google/gemma-3-1b-it, Decoding=greedy, max_new_tokens=5122026.04 | 19.6 | 1.18 | |
| GCGTarget Model=google/gemma-3-1b-it, Decoding=greedy, max_new_tokens=5122026.04 | 13.8 | 0.95 | |
| Steering-SAETarget Model=google/gemma-3-1b-it, Decoding=greedy, max_new_tokens=5122026.04 | 12.9 | 1.23 | |
| No attackTarget Model=google/gemma-3-1b-it, Decoding=greedy, max_new_tokens=5122026.04 | 12 | 1.17 | |
| PAPTarget Model=google/gemma-3-1b-it, Decoding=greedy, max_new_tokens=5122026.04 | 9.6 | 1.09 | |
| AutoDANTarget Model=google/gemma-3-1b-it, Decoding=greedy, max_new_tokens=5122026.04 | 9.3 | 0.77 |