Membership Inference Attack on C4 Vocabulary Size 140,000
0.736AUCVocabulary Overlap
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Vocabulary OverlapShadow Tokenizers=true, Auxiliary Datasets=true2025.10 | 0.736 | 69.6 | 29.72 | |
| Frequency EstimationShadow Tokenizers=true, Auxiliary Datasets=true2025.10 | 0.676 | 66 | 22.41 | |
| Naive BayesShadow Tokenizers=false, Auxiliary Datasets=true, k=60,0002025.10 | 0.543 | 53.6 | 2.8 | |
| Naive BayesShadow Tokenizers=false, Auxiliary Datasets=true, k=40,0002025.10 | 0.542 | 53.7 | 5.95 | |
| Naive BayesShadow Tokenizers=false, Auxiliary Datasets=true, k=20,0002025.10 | 0.535 | 53.3 | 4.11 | |
| Compression RateShadow Tokenizers=false, Auxiliary Datasets=false2025.10 | 0.508 | 51.4 | 0.68 | |
| Merge SimilarityShadow Tokenizers=true, Auxiliary Datasets=true2025.10 | 0.494 | 50.8 | 0.97 |