Pre-training Data Detection on ArxivMIA
0.603AUCProbe w. Synthetic Data
Evaluation Results
| Method | Links | |
|---|---|---|
| Probe w. Synthetic DataTarget Model=OpenLLaMA2024.06 | 0.603 | |
| Probe w. Real DataTarget Model=OpenLLaMA2024.06 | 0.6 | |
| Probe w. Synthetic DataTarget Model=TinyLLaMA2024.06 | 0.592 | |
| Probe w. Real DataTarget Model=TinyLLaMA2024.06 | 0.571 | |
| Neighbor AttackTarget Model=OpenLLaMA2024.06 | 0.554 | |
| Neighbor AttackTarget Model=TinyLLaMA2024.06 | 0.548 | |
| Lowercased TextTarget Model=OpenLLaMA2024.06 | 0.502 | |
| Min-K% ProbTarget Model=OpenLLaMA2024.06 | 0.492 | |
| Loss AttackTarget Model=OpenLLaMA2024.06 | 0.491 | |
| Lowercased TextTarget Model=TinyLLaMA2024.06 | 0.468 | |
| Min-K% ProbTarget Model=TinyLLaMA2024.06 | 0.455 | |
| Loss AttackTarget Model=TinyLLaMA2024.06 | 0.451 | |
| Zlib CompressionTarget Model=OpenLLaMA2024.06 | 0.438 | |
| Zlib CompressionTarget Model=TinyLLaMA2024.06 | 0.429 |