Knowledge Retrieval on Natural Questions
5.803Speedup RatioMTP-D
Evaluation results
| Method | Links | |
|---|---|---|
| MTP-DLoop strategy=4 to 16, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 5.803 | |
| MTP-DLoop strategy=4 to 8 to 16, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 5.138 | |
| MTP-DLoop strategy=4 to 8, Training protocol=Continued pre-training, Training data size=350B tokens2026.03 | 4.593 | |
| MTP-DLoop strategy=4 to 8, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 4.573 | |
| MTP-D ensembleLoop strategy=4 to 8, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 4.433 | |
| MTP-DLoop strategy=1 to 8, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 4.193 | |
| MTP-DLoop strategy=4 to 16, Training protocol=Training Free, Training data size=70B tokens2026.03 | 4.142 | |
| MTPLoop strategy=4 to 8, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 3.962 | |
| MTPLoop strategy=1 to 8, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 3.835 | |
| MTP-DLoop strategy=1 to 16, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 3.586 | |
| MTP-DLoop strategy=4 to 8, Training protocol=Training Free, Training data size=70B tokens2026.03 | 3.552 | |
| MTP-DLoop strategy=4, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 2.856 | |
| MTP-DLoop strategy=1 to 16, Training protocol=Training Free, Training data size=70B tokens2026.03 | 2.452 | |
| MTP-DLoop strategy=1 to 8, Training protocol=Training Free, Training data size=70B tokens2026.03 | 2.356 | |
| MTP-DLoop strategy=1, Training protocol=Continued pre-training, Training data size=70B tokens2026.03 | 1.212 |