Membership Inference on ORG entity category OpenLLaMA-7B (test)
83.8Balanced Accuracymultirole interrogation
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| multirole interrogationInference Target=Entity-level, Classifier=LR2026.06 | 83.8 | 88.9 | 20 | 77 | |
| multirole interrogationInference Target=Entity-level, Classifier=MV2026.06 | 83.5 | 88.6 | 16 | 72 | |
| multiangle interrogationInference Target=Entity-level, Classifier=RF2026.06 | 83 | 89.2 | 10 | 63.5 | |
| repeated interrogationInference Target=Entity-level, Classifier=RF2026.06 | 82.8 | 87.7 | 5 | 56 | |
| multiangle interrogationInference Target=Entity-level, Classifier=LR2026.06 | 82.8 | 88.6 | 8.5 | 61.5 | |
| multirole interrogationInference Target=Entity-level, Classifier=MLP2026.06 | 82.5 | 88.9 | 26.5 | 74 | |
| multiangle interrogationInference Target=Entity-level, Classifier=MV2026.06 | 82.5 | 88.3 | 6 | 58.5 | |
| multirole interrogationInference Target=Entity-level, Classifier=RF2026.06 | 82.3 | 88.5 | 26 | 73 | |
| multiangle interrogationInference Target=Entity-level, Classifier=MLP2026.06 | 82 | 88.6 | 10.5 | 62.5 | |
| repeated interrogationInference Target=Entity-level, Classifier=MLP2026.06 | 81 | 85.9 | 1.5 | 56.5 | |
| repeated interrogationInference Target=Entity-level, Classifier=LR2026.06 | 80.8 | 86.1 | 2.5 | 60.5 | |
| repeated interrogationInference Target=Entity-level, Classifier=MV2026.06 | 80.5 | 86.1 | 2.5 | 61 | |
| gradual interrogationInference Target=Entity-level, Classifier=RF2026.06 | 77.5 | 82.6 | 3.5 | 45.5 | |
| gradual interrogationInference Target=Entity-level, Classifier=MV2026.06 | 75 | 80.7 | 6.5 | 49 | |
| gradual interrogationInference Target=Entity-level, Classifier=MLP2026.06 | 75 | 81.4 | 3.5 | 46.5 | |
| gradual interrogationInference Target=Entity-level, Classifier=LR2026.06 | 74 | 80.7 | 6 | 46 | |
| misleading interrogationInference Target=Entity-level, Classifier=RF2026.06 | 73.3 | 80.1 | 16 | 42.5 | |
| misleading interrogationInference Target=Entity-level, Classifier=MLP2026.06 | 72.5 | 75.5 | 0 | 23.5 | |
| misleading interrogationInference Target=Entity-level, Classifier=MV2026.06 | 72.3 | 76 | 3 | 27.5 | |
| misleading interrogationInference Target=Entity-level, Classifier=LR2026.06 | 71.3 | 76.6 | 5 | 32 | |
| PETALInference Target=Sample-level2026.06 | 67.3 | 68.5 | 19 | 39 | |
| ICP-MIA-SP (Generation-based)Inference Target=Sample-level2026.06 | 65 | 70.8 | 8.5 | 37 | |
| ICP-MIA-SP (Masking-based)Inference Target=Sample-level2026.06 | 55.8 | 54.8 | 0.5 | 5 |