List your company

Jailbreaking MLLMs on AdvBench

100WASRwide-net-casting jailbreak method
25.22444.63764.0583.463May 16, 2026

Evaluation results

MethodLinks
wide-net-casting jailbreak methodSafety Alignment=Original Safety Alignment2026.051000.94
Ours (Wide-Net-Casting)Safety Alignment Strategy=Original Safety Alignment2026.0599.60.932
Naive Strategy 2Safety Alignment=Original Safety Alignment2026.0595.80.898
Naive Strategy 1Safety Alignment=Original Safety Alignment2026.0595.50.883
Naive Strategy 2Safety Alignment Strategy=Original Safety Alignment2026.0593.60.897
MLAI+PixArt-αSafety Alignment=Original Safety Alignment2026.0593.30.867
Naive Strategy 1Safety Alignment Strategy=Original Safety Alignment2026.0592.40.886
Baseline (MLAI + PixArt-α)Safety Alignment Strategy=Original Safety Alignment2026.0589.40.858
wide-net-casting jailbreak methodSafety Alignment=Original Safety Alignment + VLGuard2026.0550.80.473
wide-net-casting jailbreak methodSafety Alignment=Original Safety Alignment + IMMUNE2026.0547.80.44
Ours (Wide-Net-Casting)Safety Alignment Strategy=Original Safety Alignment + VLGuard2026.0546.20.412
Ours (Wide-Net-Casting)Safety Alignment Strategy=Original Safety Alignment + IMMUNE2026.0544.20.406
wide-net-casting jailbreak methodSafety Alignment=Original Safety Alignment + ASTRA2026.05420.387
Naive Strategy 2Safety Alignment=Original Safety Alignment + VLGuard2026.0541.10.363
Ours (Wide-Net-Casting)Safety Alignment Strategy=Original Safety Alignment + ASTRA2026.0540.60.365
Naive Strategy 1Safety Alignment=Original Safety Alignment + VLGuard2026.0540.60.355
Naive Strategy 2Safety Alignment=Original Safety Alignment + IMMUNE2026.0539.40.351
Naive Strategy 1Safety Alignment=Original Safety Alignment + IMMUNE2026.0538.60.344
MLAI+PixArt-αSafety Alignment=Original Safety Alignment + VLGuard2026.0537.50.311
MLAI+PixArt-αSafety Alignment=Original Safety Alignment + IMMUNE2026.0536.90.32
Naive Strategy 2Safety Alignment Strategy=Original Safety Alignment + VLGuard2026.0535.50.318
Naive Strategy 1Safety Alignment Strategy=Original Safety Alignment + VLGuard2026.0534.70.317
Naive Strategy 2Safety Alignment Strategy=Original Safety Alignment + IMMUNE2026.0534.40.307
Naive Strategy 2Safety Alignment=Original Safety Alignment + ASTRA2026.0533.90.291
Naive Strategy 1Safety Alignment Strategy=Original Safety Alignment + IMMUNE2026.0533.80.295
Naive Strategy 1Safety Alignment=Original Safety Alignment + ASTRA2026.0533.20.277
Naive Strategy 2Safety Alignment Strategy=Original Safety Alignment + ASTRA2026.0532.70.292
Baseline (MLAI + PixArt-α)Safety Alignment Strategy=Original Safety Alignment + VLGuard2026.0532.30.284
Naive Strategy 1Safety Alignment Strategy=Original Safety Alignment + ASTRA2026.05320.283
Baseline (MLAI + PixArt-α)Safety Alignment Strategy=Original Safety Alignment + IMMUNE2026.0531.50.273
MLAI+PixArt-αSafety Alignment=Original Safety Alignment + ASTRA2026.0530.70.253
Baseline (MLAI + PixArt-α)Safety Alignment Strategy=Original Safety Alignment + ASTRA2026.0528.10.242