Share your thoughts, 1 month free Claude Pro on usSee more

Home/Benchmarks

Adversarial Robustness on PGDJ64

8.46ASR

CARE

Updated 2mo ago

Evaluation Results

Method	Links
CARE 2026.03		8.46
ShiftDC 2026.03		19.73
ProFS 2026.03		31.2

SOTA Paper

CARE

Diagnosing and Repairing Unsafe Channels in Vision-Language Models via Causal Discovery and Dual-Modal Safety Subspace Projection

Dataset

PGDJ64

Follow for update

@wizwand_team Discord

Related Benchmarks

Safety Evaluation on MI-FGSM-Toxic Safety Evaluation on MI-FGSM-JailBreak Jailbreak Defense on JailBreakV Toxicity Defense on PGD-Toxic Safety Evaluation on APGD-Toxic

© 2026 wizwand

Blog Contact Changelog Swarm

Privacy Policy Terms of Service FAQs Swarm Docs