List your company
ResearchFormerly WizWand

Attribute Inference on Synthetic Dataset

71.24Attribute Inference AccuracyNo Defense
-2.849616.385235.6254.8548Feb 12, 2026
Updated 8mo ago

Evaluation results

MethodLinks
No DefenseInference Model=GPT-4o2026.0271.24—
No DefenseInference Model=Gemini 2.5 Pro2026.0268.38—
No DefenseInference Model=GPT-3.52026.0267.62—
D-DefenseInference Model=GPT-4o2026.0264.95—
D-DefenseInference Model=Gemini 2.5 Pro2026.0263.24—
D-DefenseInference Model=GPT-3.52026.0262.29—
No DefenseTarget Model=Qwen2.5-7B-Instruct2026.0257.52—
No DefenseInference Model=Llama3.1-8B-Instruct2026.0257.14—
No DefenseInference Model=Llama2-13B-Chat2026.0256.19—
AzureInference Model=GPT-4o2026.0256—
No DefenseInference Model=Llama2-7B-Chat2026.0253.71—
D-DefenseInference Model=Llama2-13B-Chat2026.0252.76—
No DefenseInference Model=DeepSeek-R1 Distill2026.0249.71—
D-DefenseInference Model=Llama3.1-8B-Instruct2026.0249.71—
D-DefenseInference Model=Llama2-7B-Chat2026.0248—
D-DefenseTarget Model=Qwen2.5-7B-Instruct2026.0248—
D-DefenseInference Model=DeepSeek-R1 Distill2026.0247.05—
Dou-SDInference Model=GPT-4o2026.0247—
FgAATarget Model=Qwen2.5-7B-Instruct, Anonymization Model=GPT-3.5-Turbo2026.0228.76—
FgAAAnonymization Model=GPT-3.5-Turbo, Inference Model=DeepSeek-R1 Distill2026.0223.24—
TRACETarget Model=Qwen2.5-7B-Instruct, Anonymization Model=GPT-3.5-Turbo2026.0222.67—
TRACEAnonymization Model=GPT-3.5-Turbo, Inference Model=Llama2-7B-Chat2026.0222.48—
FgAATarget Model=Qwen2.5-7B-Instruct, Anonymization Model=GPT-4o2026.0222.1—
TRACETarget Model=Qwen2.5-7B-Instruct, Anonymization Model=GPT-4o2026.0218.86—
MPSTarget Model=Qwen2.5-7B-Instruct2026.0210.29—
RPSInference Model=DeepSeek-R1 Distill2026.026.4817.59
TRACE-MPSTarget Model=Qwen2.5-7B-Instruct, Anonymization Model=GPT-3.5-Turbo2026.025.14—
RPSInference Model=Llama2-13B-Chat2026.024.3816.09
TRACE-MPSTarget Model=Qwen2.5-7B-Instruct, Anonymization Model=GPT-4o2026.024.38—
TRACE-RPSAnonymization Model=GPT-4o, Inference Model=DeepSeek-R1 Distill2026.022.8614.07
RPSInference Model=Llama2-7B-Chat2026.021.7114.1
TRACE-RPSAnonymization Model=GPT-3.5-Turbo, Inference Model=Llama2-7B-Chat2026.020.1912.82
RPSInference Model=Llama3.1-8B-Instruct2026.02013.48