Membership Inference Attack on C4 Vocabulary Size 200,000
0.771AUCVocabulary Overlap
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Vocabulary OverlapShadow Tokenizers=true, Auxiliary Datasets=true2025.10 | 0.771 | 71.1 | 34.61 | |
| Frequency EstimationShadow Tokenizers=true, Auxiliary Datasets=true2025.10 | 0.74 | 70.5 | 27.88 | |
| Naive BayesShadow Tokenizers=false, Auxiliary Datasets=true, k=40,0002025.10 | 0.572 | 55.7 | 10.7 | |
| Naive BayesShadow Tokenizers=false, Auxiliary Datasets=true, k=60,0002025.10 | 0.572 | 55.7 | 8.86 | |
| Naive BayesShadow Tokenizers=false, Auxiliary Datasets=true, k=20,0002025.10 | 0.564 | 55.1 | 8.03 | |
| Compression RateShadow Tokenizers=false, Auxiliary Datasets=false2025.10 | 0.509 | 51.6 | 0.77 | |
| Merge SimilarityShadow Tokenizers=true, Auxiliary Datasets=true2025.10 | 0.495 | 50.8 | 0.87 |