List your company

Transferable Adversarial Attack on HarmBench Classifier (test)

88.6TASR@1SEMA
5.08826.76948.4570.131Feb 6, 2026

Evaluation results

MethodLinks
SEMAVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0288.6
SEMAVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0287.6
CrescendoVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0281.5
SEMAVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0278
CrescendoVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0276.9
ActorAttackVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0275
ADV-LLMVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Single-turn2026.0271.7
X-TeamingVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0268.6
CoAVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0268.4
CoAVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0267.9
GOATVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0266.7
ActorAttackVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0266.7
Jigsaw PuzzleVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0265.4
Jigsaw PuzzleVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0264.3
FITDVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0263.2
DPOVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0262.8
X-TeamingVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0262.5
SFTVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0260.6
CrescendoVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0260
FITDVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0260
FITDVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0260
SFTVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0252.3
ActorAttackVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0250
DPOVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0250
GOATVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.0242.9
Jailbreak-R1Victim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.0241.2
Jailbreak-R1Victim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Single-turn2026.0240.8
CoAVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0239.3
SFTVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0238.6
ADV-LLMVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.0238.2
DPOVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0236.3
ADV-LLMVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.0234
FlipAttackVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.0233.3
Jigsaw PuzzleVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0232.1
Jailbreak-R1Victim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.0226.5
X-TeamingVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.0226.4
GOATVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.028.3