List your company
ResearchFormerly WizWand

Attack Success Rate on HEx-PHI

0Attack Success RateDataShield
-3.786821.774147.33572.8959Oct 17, 2025Nov 24, 2025Jan 2, 2026Feb 10, 2026Mar 20, 2026Apr 28, 2026Jun 6, 2026
Updated 4mo ago

Evaluation results

MethodLinks
DataShieldModel=Llama3.1, Fine-tuning Dataset=Alpaca2026.050
LARFModel=Qwen2.5, Fine-tuning Dataset=Alpaca2026.050
DataShieldModel=Qwen2.5, Fine-tuning Dataset=Alpaca2026.050
LARFModel=Llama3, Fine-tuning Dataset=Alpaca2026.050.34
DataShieldModel=Llama3, Fine-tuning Dataset=Alpaca2026.050.34
Bi-AnchorModel=Llama3.1, Fine-tuning Dataset=Alpaca2026.050.69
LARFModel=Llama3.1, Fine-tuning Dataset=Alpaca2026.050.75
Bi-AnchorModel=Llama3, Fine-tuning Dataset=Alpaca2026.051.03
DataShieldModel=Llama3, Fine-tuning Dataset=Dolly2026.051.38
DataShieldModel=Llama3.1, Fine-tuning Dataset=Dolly2026.051.38
LARFModel=Llama3.1, Fine-tuning Dataset=Dolly2026.052.07
Base2025.103
LARFModel=Llama3, Fine-tuning Dataset=Dolly2026.053.1
Bi-AnchorModel=Qwen2.5, Fine-tuning Dataset=Alpaca2026.053.45
GuardModel=Llama3.1, Fine-tuning Dataset=Alpaca2026.054.14
GuardModel=Llama3, Fine-tuning Dataset=Alpaca2026.054.83
RandomModel=Llama3.1, Fine-tuning Dataset=Alpaca2026.055.17
RandomModel=Llama3.1, Fine-tuning Dataset=Dolly2026.055.17
InstructModel=Llama3, Fine-tuning Dataset=Alpaca2026.055.86
InstructModel=Llama3, Fine-tuning Dataset=Dolly2026.055.86
SEALModel=Qwen2.5, Fine-tuning Dataset=Alpaca2026.056.55
SEALModel=Llama3, Fine-tuning Dataset=Alpaca2026.056.9
SEALModel=Llama3.1, Fine-tuning Dataset=Alpaca2026.056.9
SEALModel=Llama3, Fine-tuning Dataset=Dolly2026.057.24
LARFModel=Qwen2.5, Fine-tuning Dataset=Dolly2026.057.59
InstructModel=Llama3.1, Fine-tuning Dataset=Alpaca2026.057.93
InstructModel=Llama3.1, Fine-tuning Dataset=Dolly2026.057.93
DataShieldModel=Qwen2.5, Fine-tuning Dataset=Dolly2026.058.28
RandomModel=Qwen2.5, Fine-tuning Dataset=Alpaca2026.059.31
RandomModel=Qwen2.5, Fine-tuning Dataset=Dolly2026.059.31
RandomModel=Llama3, Fine-tuning Dataset=Alpaca2026.059.66
RandomModel=Llama3, Fine-tuning Dataset=Dolly2026.059.66
InstructModel=Qwen2.5, Fine-tuning Dataset=Alpaca2026.0510.69
InstructModel=Qwen2.5, Fine-tuning Dataset=Dolly2026.0510.69
Bi-AnchorModel=Llama3.1, Fine-tuning Dataset=Dolly2026.0512.07
GuardModel=Qwen2.5, Fine-tuning Dataset=Alpaca2026.0512.07
GradSafeModel=Llama3, Fine-tuning Dataset=Alpaca2026.0513.45
GradSafeModel=Qwen2.5, Fine-tuning Dataset=Alpaca2026.0513.79
GradSafeModel=Llama3.1, Fine-tuning Dataset=Dolly2026.0515.52
GuardModel=Llama3.1, Fine-tuning Dataset=Dolly2026.0523.45
Bi-AnchorModel=Qwen2.5, Fine-tuning Dataset=Dolly2026.0524.48
GuardModel=Qwen2.5, Fine-tuning Dataset=Dolly2026.0524.83
PatcherBase Model=Qwen2.5-1.5B2026.0626.5
GradSafeModel=Llama3.1, Fine-tuning Dataset=Alpaca2026.0529.31
PatcherBase Model=Qwen3-4B2026.0631
Bi-AnchorModel=Llama3, Fine-tuning Dataset=Dolly2026.0534.83
GradSafeModel=Qwen2.5, Fine-tuning Dataset=Dolly2026.0536.55
SEALModel=Llama3.1, Fine-tuning Dataset=Dolly2026.0538.62
GradSafeModel=Llama3, Fine-tuning Dataset=Dolly2026.0542.07
GuardModel=Llama3, Fine-tuning Dataset=Dolly2026.0544.48
SEALModel=Qwen2.5, Fine-tuning Dataset=Dolly2026.0551.03
PatcherBase Model=Llama3-8B2026.0655.5
SFTBase Model=Qwen2.5-1.5B2026.0658.6
BoosterBase Model=Qwen3-4B2026.0663.3
BoosterBase Model=Qwen2.5-1.5B2026.0663.9
VaccineBase Model=Qwen3-4B2026.0666.3
SFTBase Model=Qwen3-4B2026.0667.2
VaccineBase Model=Qwen2.5-1.5B2026.0671.7
SFT2025.1079.33
SFTBase Model=Llama3-8B2026.0691.3
VaccineBase Model=Llama3-8B2026.0692.8
BoosterBase Model=Llama3-8B2026.0693
RLVR2025.1094.67