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

Prompt Injection Defense on Jailbreak APGD

1ASR

Gradient-guided Token Suppression

-2.8423.084974.92May 24, 2026
Updated 8d ago

Evaluation Results

MethodLinks
2026.05
1
2026.05
6
2026.05
6
2026.05
6
2026.05
6
2026.05
13
2026.05
16
2026.05
27
2026.05
29
2026.05
32
2026.05
80
2026.05
97