Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Harmful Request Defense on authority probes
Loading...
75
Block Rate
Cognitive Firewall
-3
17.25
37.5
57.75
Jul 1, 2026
Block Rate
Updated 23d ago
Evaluation Results
Method
Method
Links
Block Rate
Cognitive Firewall
Defense Type=Trajector...
2026.07
75
Granite Guardian 8B
Defense Type=Per-messa...
2026.07
56
THRD-style
Defense Type=Trajector...
2026.07
31
TCA-style
Defense Type=Trajector...
2026.07
31
ShieldGemma 9B
Defense Type=Per-messa...
2026.07
12
Llama Guard 3 8B
Defense Type=Per-messa...
2026.07
6
none (undefended)
Defense Type=undefende...
2026.07
0
Feedback
Search any
task
Search any
task