Refusal Suppression on JailbreakV-28K
99.12TCRLoMC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LoMCBase model=Qwen3-VL-30B-A3B2026.06 | 99.12 | 72.14 | |
| HereticBase model=Qwen3-VL-30B-A3B2026.06 | 98.5 | 71.59 | |
| LoMCBase model=DeepSeek-VL2-small2026.06 | 98.5 | 57.63 | |
| LoMCBase model=Kimi-VL-A3B2026.06 | 98.38 | 60.36 | |
| LoMCBase model=InternVL3.5-30B-A3B2026.06 | 97.62 | 66.54 | |
| HereticBase model=DeepSeek-VL2-small2026.06 | 93.75 | 57.15 | |
| HereticBase model=InternVL3.5-30B-A3B2026.06 | 90.75 | 66.23 | |
| GateBreakerBase model=DeepSeek-VL2-small2026.06 | 77 | 54.89 | |
| SOMBase model=Kimi-VL-A3B2026.06 | 73.2 | 61.46 | |
| GateBreakerBase model=Qwen3-VL-30B-A3B2026.06 | 70 | 73.27 | |
| SOMBase model=DeepSeek-VL2-small2026.06 | 69.07 | 59.59 | |
| SOMBase model=InternVL3.5-30B-A3B2026.06 | 57.9 | 68.02 | |
| SOMBase model=Qwen3-VL-30B-A3B2026.06 | 49.07 | 69 | |
| GateBreakerBase model=InternVL3.5-30B-A3B2026.06 | 36 | 62.76 | |
| GateBreakerBase model=Kimi-VL-A3B2026.06 | 36 | 53.72 | |
| HereticBase model=Kimi-VL-A3B2026.06 | 24.25 | 59.11 | |
| OriginalBase model=Kimi-VL-A3B2026.06 | 19.54 | 60.15 | |
| OriginalBase model=DeepSeek-VL2-small2026.06 | 16.67 | 57.03 | |
| OriginalBase model=InternVL3.5-30B-A3B2026.06 | 10.06 | 66.12 | |
| OriginalBase model=Qwen3-VL-30B-A3B2026.06 | 6.32 | 71.79 |