List your company

Safety Alignment on Refusal Evaluation Dataset

99Refusal RateBase
-3.9622.7749.576.23Jan 22, 2026

Evaluation results

MethodLinks
BaseDonor=MINISTRAL-3-3B, Target=QWEN3-VL-8B, Type=REVERSE2026.0199—
BaseDonor=QWEN3-VL-2B, Target=QWEN3-VL-8B, Type=IN-FAM2026.0199—
BaseDonor=QWEN3-VL-2B, Target=MINISTRAL-3-14B, Type=CROSS-FAM2026.0198—
BaseDonor=MINISTRAL-3-3B, Target=MINISTRAL-3-14B, Type=IN-FAM2026.0198—
BaseDonor=QWEN3-VL-8B, Target=GLM-4.6V, Type=DENSE->REAS2026.0197—
BaseDonor=QWEN3-VL-8B, Target=MINISTRAL-3-3B, Type=CROSS-FAM2026.0195—
BaseDonor=QWEN3-VL-8B, Target=GPT-OSS-20B, Type=DENSE->MOE2026.0194—
BaseDonor=MINISTRAL-3-14B, Target=QWEN3-VL-2B, Type=REVERSE2026.0192—
Trajectory ReplayDonor=QWEN3-VL-8B, Target=MINISTRAL-3-3B, Type=CROSS-FAM2026.01140.81
Trajectory ReplayDonor=QWEN3-VL-8B, Target=GPT-OSS-20B, Type=DENSE->MOE2026.0180.88
Trajectory ReplayDonor=MINISTRAL-3-3B, Target=QWEN3-VL-8B, Type=REVERSE2026.0150.92
Trajectory ReplayDonor=QWEN3-VL-2B, Target=MINISTRAL-3-14B, Type=CROSS-FAM2026.0120.96
Trajectory ReplayDonor=MINISTRAL-3-14B, Target=QWEN3-VL-2B, Type=REVERSE2026.0110.98
Trajectory ReplayDonor=QWEN3-VL-8B, Target=GLM-4.6V, Type=DENSE->REAS2026.0110.98
Trajectory ReplayDonor=QWEN3-VL-2B, Target=QWEN3-VL-8B, Type=IN-FAM2026.0100.99
Trajectory ReplayDonor=MINISTRAL-3-3B, Target=MINISTRAL-3-14B, Type=IN-FAM2026.0100.99