List your company

Jailbreak Defense on TAP

2ASRWildGuard (Pre)
-119.2539.559.75Apr 1, 2026Apr 2, 2026Apr 3, 2026Apr 4, 2026Apr 5, 2026Apr 6, 2026Apr 8, 2026

Evaluation results

MethodLinks
WildGuard (Pre)2026.0428
GuardReasoner (Pre)Latency (Sec.)=14.87, Memory Overhead (MB)=15617.332026.0437
WildGuard (Post)2026.04487
GradientCuffLatency (Sec.)=26.57, Memory Overhead (MB)=2063.762026.04510
Token Highlighter2026.0457
GuardReasoner (Post)Latency (Sec.)=16.71, Memory Overhead (MB)=15618.132026.04587
SelfDefend (Intent)Latency (Sec.)=3.11, Memory Overhead (MB)=13457.402026.04612
SelfGraderLatency (Sec.)=1.24, Memory Overhead (MB)=479.812026.04620
GradSafe2026.04953
SelfDefend (Direct)Latency (Sec.)=1.19, Memory Overhead (MB)=13452.932026.041220
LLama-3-8B-Instruct (No Defense)Latency (Sec.)=1.472026.0414—
Perplexity FilterLatency (Sec.)=0.49, Memory Overhead (MB)=13571.942026.0414100
Prompt GuardLatency (Sec.)=22.67, Memory Overhead (MB)=1864.112026.041493
Llama Guard (Pre)Latency (Sec.)=0.96, Memory Overhead (MB)=13910.612026.041446
Llama Guard (Post)Latency (Sec.)=1.17, Memory Overhead (MB)=13910.612026.0414100
FedDetoxSetting=Ours2026.0461—
Qwen2.5-1.5B-InstructSetting=Original2026.0467—
FedDPOSetting=Poisoned2026.0477—