List your company

Jailbreak Attack on JailbreakBench

0ASR@10DR
-3.5620.4744.568.53Apr 25, 2024Aug 12, 2024Nov 29, 2024Mar 18, 2025Jul 5, 2025Oct 22, 2025Feb 9, 2026

Evaluation results

MethodLinks
DRTarget Model=Llama-2-7b-chat2024.040———
HumanTarget Model=Llama-2-7b-chat2024.040———
DRTarget Model=Llama-2-13b-chat2024.040———
HumanTarget Model=Llama-3-8B-Instruct2024.040———
DRTarget Model=Gemma-2-9b-it2024.040———
OriginalTarget Model=OLMoE-1B-7B2026.020———
SHIPSTarget Model=DeepSeek-V2-Lite2026.020———
TAPTarget Model=Llama-2-7b-chat2024.041———
TAPTarget Model=Llama-2-13b-chat2024.041———
HumanTarget Model=Llama-2-13b-chat2024.041———
HumanTarget Model=Llama-3.1-8B-Instruct2024.041———
DRTarget Model=Llama-3.1-8B-Instruct2024.042———
DRTarget Model=Llama-3-8B-Instruct2024.044———
DRTarget Model=Vicuna-13b-v1.52024.044———
DRTarget Model=Qwen2.5-7B-Instruct2024.045———
PAIRTarget Model=Llama-3.1-8B-Instruct2024.046———
TAPTarget Model=Llama-3.1-8B-Instruct2024.047———
DRTarget Model=Qwen2-7B-Chat2024.047———
OriginalTarget Model=Qwen1.5-MoE-A2.7B2026.027———
SAFExTarget Model=OLMoE-1B-7B2026.027———
TAPTarget Model=Llama-3-8B-Instruct2024.048———
PAIRTarget Model=Llama-2-13b-chat2024.049———
PAIRTarget Model=Llama-2-7b-chat2024.0410———
DRTarget Model=Vicuna-7b-v1.32024.0411———
DRTarget Model=Vicuna-7b-v1.52024.0411———
PAIRTarget Model=Llama-3-8B-Instruct2024.0414———
SteerMoETarget Model=OLMoE-1B-7B2026.0214———
TAPTarget Model=OLMoE-1B-7B2026.0214———
RSTarget Model=Llama-2-7b-chat2024.0415———
OriginalTarget Model=DeepSeek-V2-Lite2026.0215———
OriginalTarget Model=Average2026.0216———
SHIPSTarget Model=OLMoE-1B-7B2026.0216———
RSTarget Model=Llama-2-13b-chat2024.0421———
PAIRTarget Model=Gemma-2-9b-it2024.0424———
GCGTarget Model=OLMoE-1B-7B2026.0224———
TAPTarget Model=Gemma-2-9b-it2024.0426———
SHIPSTarget Model=Average2026.0229———
SAFExTarget Model=DeepSeek-V2-Lite2026.0231———
SAFExTarget Model=Average2026.0231———
SHIPSTarget Model=Qwen1.5-MoE-A2.7B2026.0233———
TAPTarget Model=Qwen2.5-7B-Instruct2024.0434———
Dataset-Level Top-5 RoSaisTarget Model=OLMoE-1B-7B2026.0234———
TAPTarget Model=Mixtral-8x7B2026.0236———
Dataset-Level Top-5 RoSaisTarget Model=Qwen1.5-MoE-A2.7B2026.0237———
GCGTarget Model=DeepSeek-V2-Lite2026.0238———
DRTarget Model=Mistral-7B-Instruct-v0.22024.0439———
SAFExTarget Model=Qwen1.5-MoE-A2.7B2026.0239———
OriginalTarget Model=Mixtral-8x7B2026.0240———
TAPTarget Model=Vicuna-13b-v1.52024.0441———
PAIRTarget Model=Qwen2-7B-Chat2024.0442———
PAIRTarget Model=Qwen2.5-7B-Instruct2024.0444———
DRTarget Model=Mistral-7B-Instruct-v0.32024.0444———
Sample-Level Top-5 RoSaisTarget Model=OLMoE-1B-7B2026.0245———
GCGTarget Model=Average2026.0246———
SAFExTarget Model=Mixtral-8x7B2026.0246———
SteerMoETarget Model=Qwen1.5-MoE-A2.7B2026.0246———
Sample-Level Top-5 RoSaisTarget Model=DeepSeek-V2-Lite2026.0246———
PAIRTarget Model=Vicuna-13b-v1.52024.0447———
TAPTarget Model=Average2026.0247———
TAPTarget Model=Qwen2-7B-Chat2024.0449———
SteerMoETarget Model=Average2026.0250———
TAPTarget Model=Vicuna-7b-v1.52024.0451———
PAIRTarget Model=Mistral-7B-Instruct-v0.22024.0452———
PAIRTarget Model=Mistral-7B-Instruct-v0.32024.0452———
PAIRTarget Model=Vicuna-7b-v1.32024.0454———
TAPTarget Model=Vicuna-7b-v1.32024.0455———
Dataset-Level Top-5 RoSaisTarget Model=Average2026.0255———
TAPTarget Model=Mistral-7B-Instruct-v0.32024.0457———
GCGTarget Model=Llama-3.1-8B-Instruct2024.0458———
PAIRTarget Model=Vicuna-7b-v1.52024.0458———
GCGTarget Model=Qwen1.5-MoE-A2.7B2026.0258———
Sample-Level Top-5 RoSaisTarget Model=Average2026.0259———
TAPTarget Model=DeepSeek-V2-Lite2026.0260———
TAPTarget Model=Mistral-7B-Instruct-v0.22024.0461———
GCGTarget Model=Mixtral-8x7B2026.0263———
Sample-Level Top-5 RoSaisTarget Model=Qwen1.5-MoE-A2.7B2026.0263———
RSTarget Model=Llama-3.1-8B-Instruct2024.0464———
SteerMoETarget Model=DeepSeek-V2-Lite2026.0266———
SHIPSTarget Model=Mixtral-8x7B2026.0267———
Dataset-Level Top-5 RoSaisTarget Model=Mixtral-8x7B2026.0268———
HumanTarget Model=Qwen2.5-7B-Instruct2024.0470———
SteerMoETarget Model=Mixtral-8x7B2026.0273———
GCGTarget Model=Llama-3-8B-Instruct2024.0474———
HumanTarget Model=Qwen2-7B-Chat2024.0474———
GCGTarget Model=Llama-2-7b-chat2024.0476———
TAPTarget Model=Qwen1.5-MoE-A2.7B2026.0277———
Dataset-Level Top-5 RoSaisTarget Model=DeepSeek-V2-Lite2026.0279———
GCGTarget Model=Llama-2-13b-chat2024.0480———
DSNTarget Model=Llama-3.1-8B-Instruct2024.0481———
GCGTarget Model=Vicuna-7b-v1.32024.0481———
Sample-Level Top-5 RoSaisTarget Model=Mixtral-8x7B2026.0281———
RSTarget Model=Llama-3-8B-Instruct2024.0483———
RS_self-transferTarget Model=Llama-2-7b-chat2024.0484———
F-SOURTarget Model=OLMoE-1B-7B2026.0286———
HumanTarget Model=Vicuna-7b-v1.52024.0487———
GCGTarget Model=Gemma-2-9b-it2024.0488———
HumanTarget Model=Vicuna-7b-v1.32024.0488———
GCGTarget Model=Vicuna-7b-v1.52024.0488———
F-SOURTarget Model=Qwen1.5-MoE-A2.7B2026.0288———
RS_self-transferTarget Model=Llama-3-8B-Instruct2024.0489———