Training Data Reconstruction on CoLA
100ROUGE-1UTR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| UTRBatch Size=1, Model Architecture=BERT2026.01 | 100 | 100 | |
| UTRBatch Size=1, Model Architecture=GPT22026.01 | 100 | 100 | |
| UTRBatch Size=1, Model Architecture=Qwen2026.01 | 100 | 100 | |
| DAGERBatch Size=1, Model Architecture=BERT2026.01 | 100 | 100 | |
| DAGERBatch Size=1, Model Architecture=GPT22026.01 | 100 | 100 | |
| UTRBatch Size=2, Model Architecture=BERT2026.01 | 100 | 100 | |
| UTRBatch Size=2, Model Architecture=Qwen2026.01 | 100 | 100 | |
| DAGERBatch Size=2, Model Architecture=BERT2026.01 | 100 | 100 | |
| DAGERBatch Size=2, Model Architecture=GPT22026.01 | 100 | 100 | |
| UTRBatch Size=4, Model Architecture=BERT2026.01 | 100 | 100 | |
| UTRBatch Size=4, Model Architecture=Qwen2026.01 | 100 | 100 | |
| DAGERBatch Size=4, Model Architecture=GPT22026.01 | 100 | 100 | |
| UTRBatch Size=8, Model Architecture=BERT2026.01 | 100 | 100 | |
| UTRBatch Size=8, Model Architecture=Qwen2026.01 | 100 | 100 | |
| UTRBatch Size=4, Model Architecture=GPT22026.01 | 99.92 | 99.88 | |
| UTRBatch Size=8, Model Architecture=GPT22026.01 | 99.92 | 99.83 | |
| UTRBatch Size=2, Model Architecture=GPT22026.01 | 99.83 | 99.83 | |
| DAGERBatch Size=4, Model Architecture=BERT2026.01 | 94 | 89.9 | |
| LAMPBatch Size=1, Model Architecture=GPT22026.01 | 89.6 | 51.9 | |
| DAGERBatch Size=8, Model Architecture=GPT22026.01 | 88.18 | 84.5 | |
| LAMPBatch Size=2, Model Architecture=GPT22026.01 | 77.8 | 31.5 | |
| LAMPBatch Size=1, Model Architecture=BERT2026.01 | 73.3 | 43.3 | |
| DAGERBatch Size=8, Model Architecture=BERT2026.01 | 67.8 | 48.8 | |
| LAMPBatch Size=4, Model Architecture=GPT22026.01 | 66.2 | 21.8 | |
| LAMPBatch Size=8, Model Architecture=GPT22026.01 | 52.9 | 13.1 | |
| LAMPBatch Size=2, Model Architecture=BERT2026.01 | 26.8 | 11 | |
| LAMPBatch Size=4, Model Architecture=BERT2026.01 | 13.4 | 3.9 | |
| LAMPBatch Size=8, Model Architecture=BERT2026.01 | 8.9 | 1.9 | |
| DAGERBatch Size=1, Model Architecture=Qwen2026.01 | 0 | 0 | |
| DAGERBatch Size=2, Model Architecture=Qwen2026.01 | 0 | 0 | |
| DAGERBatch Size=4, Model Architecture=Qwen2026.01 | 0 | 0 | |
| DAGERBatch Size=8, Model Architecture=Qwen2026.01 | 0 | 0 |