List your company
ResearchFormerly WizWand

Jailbreak Defense on AdvBench PAD

12.12ASRLLaDA-1.5 + Self-reminder + DIFFUGUARD
8.604832.332456.0679.7876Sep 29, 2025
Updated 7mo ago

Evaluation results

MethodLinks
LLaDA-1.5 + Self-reminder + DIFFUGUARDBase Model=LLaDA-1.5, Defense Method=Self-reminder + DIFFUGUARD2025.0912.12
LLaDA-1.5 + Self-reminderBase Model=LLaDA-1.5, Defense Method=Self-reminder2025.0915
LLaDA + DIFFUGUARDdefense=DIFFUGUARD2025.0917.12
Dream + DIFFUGUARDdefense=DIFFUGUARD2025.0919.42
LLaDA-8B-Instruct + Self-reminder + DIFFUGUARDBase Model=LLaDA-8B-Instruct, Defense Method=Self-reminder + DIFFUGUARD2025.0924.42
LLaDA-8B w/ DIFFUGUARDAttack Scenario=Standard2025.0924.42
LLaDA-8B-Instruct + Self-reminderBase Model=LLaDA-8B-Instruct, Defense Method=Self-reminder2025.0930.58
Dream-v0-Instruct-7B + DIFFUGUARDBase Model=Dream-v0-Instruct-7B, Defense Method=DIFFUGUARD2025.0931.15
LLaDA-8B w/ DIFFUGUARDAttack Scenario=Multi-sampling2025.0935.96
Dream-v0-Instruct-7B + Self-reminder + DIFFUGUARDBase Model=Dream-v0-Instruct-7B, Defense Method=Self-reminder + DIFFUGUARD2025.0937.31
Dream-7B w/ DIFFUGUARDAttack Scenario=Standard2025.0937.31
MMaDA-8B-MixCoT + Self-reminder + DIFFUGUARDBase Model=MMaDA-8B-MixCoT, Defense Method=Self-reminder + DIFFUGUARD2025.0937.69
Dream-7B w/ DIFFUGUARDAttack Scenario=Multi-sampling2025.0940.38
LLaDA-8B w/ DIFFUGUARDAttack Scenario=Gradient-based2025.0944.42
LLaDA-1.5 + DIFFUGUARDBase Model=LLaDA-1.5, Defense Method=DIFFUGUARD2025.0956.15
LLaDA-8B-Instruct + DIFFUGUARDBase Model=LLaDA-8B-Instruct, Defense Method=DIFFUGUARD2025.0959.62
MMaDA-8B-MixCoT + DIFFUGUARDBase Model=MMaDA-8B-MixCoT, Defense Method=DIFFUGUARD2025.0961.54
LLaDA-8B w/ DIFFUGUARDAttack Scenario=Threshold-probing2025.0965.58
Dream-7B w/ DIFFUGUARDAttack Scenario=Gradient-based2025.0965.96
Dream2025.0970.38
LLaDA-1.5 + PPL-FilterBase Model=LLaDA-1.5, Defense Method=PPL-Filter2025.0974.42
MMaDA-8B-MixCoT + Self-reminderBase Model=MMaDA-8B-MixCoT, Defense Method=Self-reminder2025.0978.08
Dream-7B w/ DIFFUGUARDAttack Scenario=Threshold-probing2025.0984.62
LLaDA-8B-Instruct + PPL-FilterBase Model=LLaDA-8B-Instruct, Defense Method=PPL-Filter2025.0985.96
MMaDA-8B-MixCoT + PPL-FilterBase Model=MMaDA-8B-MixCoT, Defense Method=PPL-Filter2025.0986.15
LLaDA2025.0986.54
LLaDA-1.5Base Model=LLaDA-1.5, Defense Method=None2025.0987.69
LLaDA-8BAttack Scenario=Gradient-based2025.0988.27
LLaDA-8B-InstructBase Model=LLaDA-8B-Instruct, Defense Method=None2025.0993.65
LLaDA-8BAttack Scenario=Standard2025.0993.65
Dream-v0-Instruct-7B + PPL-FilterBase Model=Dream-v0-Instruct-7B, Defense Method=PPL-Filter2025.0994.04
Dream-7BAttack Scenario=Gradient-based2025.0995.96
Dream-7BAttack Scenario=Threshold-probing2025.0998.27
Dream-v0-Instruct-7B + Self-reminderBase Model=Dream-v0-Instruct-7B, Defense Method=Self-reminder2025.0998.65
LLaDA-8BAttack Scenario=Threshold-probing2025.0998.84
LLaDA-8BAttack Scenario=Multi-sampling2025.0998.85
Dream-v0-Instruct-7BBase Model=Dream-v0-Instruct-7B, Defense Method=None2025.0999.23
MMaDA-8B-MixCoTBase Model=MMaDA-8B-MixCoT, Defense Method=None2025.0999.23
Dream-7BAttack Scenario=Standard2025.0999.23
Dream-7BAttack Scenario=Multi-sampling2025.09100