List your company

Jailbreak Defense on Wild Jailbreak

0.1ASRDream + DIFFUGUARD
-2.35814.233530.82547.4165Nov 15, 2023Apr 5, 2024Aug 25, 2024Jan 15, 2025Jun 6, 2025Oct 26, 2025Mar 18, 2026

Evaluation results

MethodLinks
Dream + DIFFUGUARDdefense=DIFFUGUARD2025.090.1——
Dream2025.090.4——
OursModel=GPT-4-0125 (2024)2023.110.5——
Dream-v0-Instruct-7B + Self-reminder + DIFFUGUARDBase Model=Dream-v0-Instruct-7B, Defense Method=Self-reminder + DIFFUGUARD2025.091.05——
Dream-v0-Instruct-7B + Self-reminderBase Model=Dream-v0-Instruct-7B, Defense Method=Self-reminder2025.091.7——
Self-ReminderModel=GPT-4-0125 (2024)2023.112.3——
Dream-v0-Instruct-7B + DIFFUGUARDBase Model=Dream-v0-Instruct-7B, Defense Method=DIFFUGUARD2025.092.35——
LLaDA + DIFFUGUARDdefense=DIFFUGUARD2025.092.45——
OursModel=GPT-4-0613 (2023)2023.113.1——
Dream-v0-Instruct-7BBase Model=Dream-v0-Instruct-7B, Defense Method=None2025.093.3——
Dream-v0-Instruct-7B + PPL-FilterBase Model=Dream-v0-Instruct-7B, Defense Method=PPL-Filter2025.093.3——
OursModel=ChatGPT-0613 (2023)2023.113.6——
OursModel=ChatGPT-0125 (2024)2023.113.6——
Self-GuardModel Scale=Qwen3-14B2026.014.8——
ReasoningGuardModel Scale=Qwen3-14B2026.014.9——
Self-ReminderModel Scale=Qwen3-14B2026.016.3——
Self-ReminderModel=GPT-4-0613 (2023)2023.117.2——
LLaDA2025.097.75——
Self-GuardModel Scale=Qwen3-8B2026.018——
ReasoningGuardModel Scale=Qwen3-8B2026.018.5——
LLaDA-8B-Instruct + Self-reminder + DIFFUGUARDBase Model=LLaDA-8B-Instruct, Defense Method=Self-reminder + DIFFUGUARD2025.098.5——
STAR-1Model Scale=Qwen3-14B2026.018.9——
SafeKeyModel Scale=Qwen3-14B2026.019.8——
Self-ReminderModel Scale=Qwen3-8B2026.0110.4——
LLaDA-1.5 + Self-reminder + DIFFUGUARDBase Model=LLaDA-1.5, Defense Method=Self-reminder + DIFFUGUARD2025.0910.95——
VanillaModel=GPT-4-0125 (2024)2023.1111.9——
SafeKeyModel Scale=Qwen3-8B2026.0112.3——
LLaDA-1.5 + Self-reminderBase Model=LLaDA-1.5, Defense Method=Self-reminder2025.0912.65——
Self-GuardModel Scale=Qwen3-4B2026.0112.9——
ReasoningGuardModel Scale=Qwen3-4B2026.0113——
STAR-1Model Scale=Qwen3-8B2026.0113——
PreSafeBase Model=Qwen-34B2026.0314——
MMaDA-8B-MixCoT + Self-reminder + DIFFUGUARDBase Model=MMaDA-8B-MixCoT, Defense Method=Self-reminder + DIFFUGUARD2025.0914.25——
SafeKeyModel Scale=Qwen3-4B2026.0114.5——
STAR-1Model Scale=Qwen3-4B2026.0115.4——
LLaDA-8B-Instruct + Self-reminderBase Model=LLaDA-8B-Instruct, Defense Method=Self-reminder2025.0916——
Self-ReminderModel Scale=Qwen3-4B2026.0116.9——
MMaDA-8B-MixCoT + Self-reminderBase Model=MMaDA-8B-MixCoT, Defense Method=Self-reminder2025.0918.3——
PreSafeBase Model=DS-R1-7B2026.0318.8——
PreSafeBase Model=Qwen-8B2026.0318.8——
PreSafeBase Model=DS-R1-14B2026.0319.2——
Self-ReminderModel=ChatGPT-0125 (2024)2023.1120.2——
LLaDA-8B-Instruct + DIFFUGUARDBase Model=LLaDA-8B-Instruct, Defense Method=DIFFUGUARD2025.0921——
LLaDA-8B-Instruct + PPL-FilterBase Model=LLaDA-8B-Instruct, Defense Method=PPL-Filter2025.0922.75——
PreSafeBase Model=DS-R1-8B2026.0322.8——
LLaDA-8B-InstructBase Model=LLaDA-8B-Instruct, Defense Method=None2025.0923.95——
LLaDA-1.5 + DIFFUGUARDBase Model=LLaDA-1.5, Defense Method=DIFFUGUARD2025.0924.65——
PreSafeBase Model=Skywork-7B2026.0324.8——
Erase And CheckBase Model=DS-R1-8B2026.0325.2——
Erase And CheckBase Model=DS-R1-7B2026.0325.6——
Perplexity FilterBase Model=Qwen-8B2026.0326——
Erase And CheckBase Model=Skywork-7B2026.0326——
LLaDA-1.5 + PPL-FilterBase Model=LLaDA-1.5, Defense Method=PPL-Filter2025.0926.15——
Erase And CheckBase Model=Qwen-8B2026.0326.4——
Erase And CheckBase Model=Qwen-34B2026.0327.2——
LLaDA-1.5Base Model=LLaDA-1.5, Defense Method=None2025.0927.4——
Alpha-steerModel Scale=Qwen3-14B2026.0127.7——
Self-ReminderModel=ChatGPT-0613 (2023)2023.1128.1——
Erase And CheckBase Model=DS-R1-14B2026.0330.4——
R2DBase Model=Qwen-8B2026.0332.6——
InstructModel Scale=Qwen3-14B2026.0133.1——
SafeChainModel Scale=Qwen3-4B2026.0133.9——
SafeChainModel Scale=Qwen3-8B2026.0134.4——
SafeChainModel Scale=Qwen3-14B2026.0134.4——
SafeChainBase Model=Skywork-7B2026.0336.8——
InstructModel Scale=Qwen3-8B2026.0140.3——
SafeChainBase Model=DS-R1-8B2026.0340.4——
SafeChainBase Model=Qwen-8B2026.0342——
SafeChainBase Model=DS-R1-14B2026.0342.8——
R2DBase Model=DS-R1-14B2026.0342.9——
Alpha-steerModel Scale=Qwen3-4B2026.0143.9——
InstructModel Scale=Qwen3-4B2026.0144.3——
Alpha-steerModel Scale=Qwen3-8B2026.0144.6——
Perplexity FilterBase Model=DS-R1-14B2026.0346——
SafeChainBase Model=Qwen-34B2026.0346——
SafeChainBase Model=DS-R1-7B2026.0348——
VanillaModel=GPT-4-0613 (2023)2023.1148.3——
Smooth LLMBase Model=Qwen-8B2026.0349.6——
Remove Non DictionaryBase Model=Qwen-8B2026.0349.6——
Smooth LLMBase Model=DS-R1-14B2026.0350.4——
NoneBase Model=Qwen-8B2026.0351.2——
VanillaModel=ChatGPT-0125 (2024)2023.1152.1——
Remove Non DictionaryBase Model=Qwen-34B2026.0352.8——
R2DBase Model=Qwen-34B2026.0352.8——
Remove Non DictionaryBase Model=DS-R1-14B2026.0353.6——
Perplexity FilterBase Model=Qwen-34B2026.0353.6——
NoneBase Model=DS-R1-14B2026.0354——
NoneBase Model=Qwen-34B2026.0354——
Synonym SubstitutionBase Model=DS-R1-14B2026.0355.6——
Remove Non DictionaryBase Model=DS-R1-8B2026.0357.2——
Smooth LLMBase Model=DS-R1-8B2026.0357.6——
Synonym SubstitutionBase Model=Qwen-8B2026.0358.4——
NoneBase Model=DS-R1-8B2026.0359.2——
Smooth LLMBase Model=Qwen-34B2026.0359.6——
MMaDA-8B-MixCoT + PPL-FilterBase Model=MMaDA-8B-MixCoT, Defense Method=PPL-Filter2025.0959.85——
R2DBase Model=Skywork-7B2026.0360.2——
Synonym SubstitutionBase Model=Qwen-34B2026.0360.8——
Remove Non DictionaryBase Model=Skywork-7B2026.0360.8——
Synonym SubstitutionBase Model=DS-R1-8B2026.0361.2——
MMaDA-8B-MixCoT + DIFFUGUARDBase Model=MMaDA-8B-MixCoT, Defense Method=DIFFUGUARD2025.0961.55——