Membership Inference on Llama-8B membership dataset
95.7AUCCRM-LTS
Evaluation results
| Method | Links | |
|---|---|---|
| CRM-LTSFeatures=per-layer (32-48 dims), Protocol=5-fold CV, Sample size=250 samples2026.05 | 95.7 | |
| Logit LensFeatures=per-layer (32-48 dims), Protocol=5-fold CV, Sample size=250 samples2026.05 | 83.1 | |
| Attention FlowFeatures=per-layer (32-48 dims), Protocol=5-fold CV, Sample size=250 samples2026.05 | 78.4 | |
| Gradient NormFeatures=per-layer (32-48 dims), Protocol=5-fold CV, Sample size=250 samples2026.05 | 50 |