List your company
ResearchFormerly WizWand

Transferable Adversarial Attack on AdvBench LLM Classifier (test)

9,260TASR@1SEMA
-370.42,129.84,6307,130.2Feb 6, 2026
Updated 8mo ago

Evaluation results

MethodLinks
SEMAVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.029,260
SEMAVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.029,110
SEMAVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.028,510
CrescendoVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.027,860
CrescendoVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.027,700
ActorAttackVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.027,110
ADV-LLMVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Single-turn2026.027,010
ActorAttackVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,810
FITDVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,790
GOATVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,780
X-TeamingVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,750
X-TeamingVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,630
GOATVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,590
FITDVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,540
CoAVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,320
SFTVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,290
DPOVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,100
FITDVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.026,060
CoAVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.026,030
Jigsaw PuzzleVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.025,860
SFTVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.025,800
ActorAttackVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.025,560
Jigsaw PuzzleVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.025,550
CrescendoVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.025,510
DPOVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Multi-turn2026.025,240
Jigsaw PuzzleVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.024,790
CoAVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.024,660
SFTVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.024,450
Jailbreak-R1Victim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.024,400
DPOVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.023,810
X-TeamingVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.023,760
Jailbreak-R1Victim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Single-turn2026.023,670
FlipAttackVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.023,330
Jailbreak-R1Victim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.023,170
GOATVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Multi-turn2026.022,100
FlipAttackVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.021,110
ADV-LLMVictim (Source)=Llama-3.1-8B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.021,000
ADV-LLMVictim (Source)=Qwen2.5-3B, Victim (Target)=GPT-4.1-mini, Turns=Single-turn2026.02790
FlipAttackVictim (Source)=Qwen2.5-3B, Victim (Target)=Llama-3.1-8B, Turns=Single-turn2026.020