List your company

Refusal Suppression on AdvBench

3.33Targeted Content Rate (TCR)Original
-0.436824.989150.41575.8409Jun 10, 2026

Evaluation results

MethodLinks
OriginalBase model=Kimi-VL-A3B2026.063.3360.15
OriginalBase model=DeepSeek-VL2-small2026.06557.03
OriginalBase model=Qwen3-VL-30B-A3B2026.0611.3271.79
SOMBase model=Qwen3-VL-30B-A3B2026.0616.9868.96
OriginalBase model=InternVL3.5-30B-A3B2026.0618.2466.12
GateBreakerBase model=Kimi-VL-A3B2026.0630.8249.35
SOMBase model=Kimi-VL-A3B2026.0648.4360.16
HereticBase model=InternVL3.5-30B-A3B2026.0655.8366.29
GateBreakerBase model=InternVL3.5-30B-A3B2026.0657.2363.1
HereticBase model=Qwen3-VL-30B-A3B2026.0660.8371.75
GateBreakerBase model=Qwen3-VL-30B-A3B2026.0667.9273.21
SOMBase model=DeepSeek-VL2-small2026.0669.1857.97
GateBreakerBase model=DeepSeek-VL2-small2026.0669.1857.21
HereticBase model=DeepSeek-VL2-small2026.067556.45
SOMBase model=InternVL3.5-30B-A3B2026.0682.3968.11
HereticBase model=Kimi-VL-A3B2026.068559.03
LoMCBase model=Qwen3-VL-30B-A3B2026.0689.1772.23
LoMCBase model=DeepSeek-VL2-small2026.069057.75
LoMCBase model=Kimi-VL-A3B2026.0693.3360.28
LoMCBase model=InternVL3.5-30B-A3B2026.0697.566.46