List your company
ResearchFormerly WizWand

Defense against Harmful Fine-tuning on Backdoor Jailbreaking No Trigger

1.6Harm ScoreBooster
1.5361.9682.42.832May 7, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Booster2026.051.6
SBR2026.051.8
Lisa2026.052.1
Vaccine2026.052.3
DeepAlign2026.052.7
SFT2026.053.2