Pre-training Data Detection on ArxivMIA CS
67.3AUCProbe w. Synthetic Data
Evaluation Results
| Method | Links | |
|---|---|---|
| Probe w. Synthetic DataTarget Model=OpenLLaMA2024.06 | 67.3 | |
| Probe w. Real DataTarget Model=OpenLLaMA2024.06 | 67.2 | |
| Probe w. Synthetic DataTarget Model=TinyLLaMA2024.06 | 64.3 | |
| Probe w. Real DataTarget Model=TinyLLaMA2024.06 | 63.7 | |
| Neighbor AttackTarget Model=TinyLLaMA2024.06 | 59.3 | |
| Neighbor AttackTarget Model=OpenLLaMA2024.06 | 59.3 | |
| Smaller ModelTarget Model=TinyLLaMA2024.06 | 55.9 | |
| Smaller ModelTarget Model=OpenLLaMA2024.06 | 54.9 | |
| Loss AttackTarget Model=OpenLLaMA2024.06 | 51.4 | |
| Min-K% ProbTarget Model=OpenLLaMA2024.06 | 50.2 | |
| Lowercased TextTarget Model=OpenLLaMA2024.06 | 47.8 | |
| Loss AttackTarget Model=TinyLLaMA2024.06 | 45.3 | |
| Min-K% ProbTarget Model=TinyLLaMA2024.06 | 45 | |
| Lowercased TextTarget Model=TinyLLaMA2024.06 | 43.8 | |
| Zlib CompressionTarget Model=OpenLLaMA2024.06 | 40.4 | |
| Zlib CompressionTarget Model=TinyLLaMA2024.06 | 38 |