Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Adversarial Attack on HarmBench standard behaviors
Loading...
91
ASR
AdvGRPO
8.32
29.785
51.25
72.715
Jun 8, 2026
ASR
Updated 1mo ago
Evaluation Results
Method
Method
Links
ASR
AdvGRPO
Attacker=Qwen2.5-14B,...
2026.06
91
AdvGRPO
Attacker=Qwen2.5-7B, D...
2026.06
84.5
SEMA
Attacker=Qwen2.5-14B,...
2026.06
84.5
SEMA
Attacker=Qwen2.5-3B, D...
2026.06
80.5
AdvGRPO
Attacker=Qwen2.5-14B,...
2026.06
79.5
AdvGRPO
Attacker=Qwen3.5-9B, D...
2026.06
71
AdvGRPO
Attacker=Qwen2.5-7B, D...
2026.06
69.5
GRP-Oblit
Attacker=Qwen-2.5-14B,...
2026.06
55
Unsafe-SFT
Attacker=GPT-4o, Defen...
2026.06
35.5
GRP-Oblit
Attacker=Qwen-2.5-14B,...
2026.06
25.5
Abliteration
Attacker=Llama-3.1-8B,...
2026.06
14
Abliteration
Attacker=Llama-3.1-8B,...
2026.06
13.5
Unsafe-SFT
Attacker=GPT-4o, Defen...
2026.06
11.5
Feedback
Search any
task
Search any
task