Multi-sensor Vision-Language Modeling on MAPLM
58.6FRM ScoreDriveMM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DriveMMBase Model=DriveMM, Retained Tokens=729, Pruning Ratio=0%2026.06 | 58.6 | 85.87 | 72.24 | |
| MVPrunerBase Model=DriveMM, Retained Tokens=180, Pruning Ratio=75%2026.06 | 56.07 | 84.55 | 70.31 | |
| Prune2DriveBase Model=DriveMM, Retained Tokens=180, Pruning Ratio=75%2026.06 | 53.47 | 83.1 | 68.29 | |
| DivPruneBase Model=DriveMM, Retained Tokens=180, Pruning Ratio=75%2026.06 | 53.46 | 83.1 | 68.28 | |
| MVPrunerBase Model=DriveMM, Retained Tokens=72, Pruning Ratio=90%2026.06 | 53.27 | 83.33 | 68.3 | |
| DARTBase Model=DriveMM, Retained Tokens=180, Pruning Ratio=75%2026.06 | 51.27 | 83.27 | 67.27 | |
| PACTBase Model=DriveMM, Retained Tokens=180, Pruning Ratio=75%2026.06 | 50.87 | 82.5 | 66.69 | |
| SparseVLMBase Model=DriveMM, Retained Tokens=180, Pruning Ratio=75%2026.06 | 49.8 | 82.52 | 66.16 | |
| DivPruneBase Model=DriveMM, Retained Tokens=72, Pruning Ratio=90%2026.06 | 48.8 | 80.75 | 64.78 | |
| Prune2DriveBase Model=DriveMM, Retained Tokens=72, Pruning Ratio=90%2026.06 | 48.33 | 81.28 | 64.81 | |
| FastVBase Model=DriveMM, Retained Tokens=180, Pruning Ratio=75%2026.06 | 47.67 | 81.7 | 64.69 | |
| DARTBase Model=DriveMM, Retained Tokens=72, Pruning Ratio=90%2026.06 | 40.27 | 79.23 | 59.75 | |
| SparseVLMBase Model=DriveMM, Retained Tokens=72, Pruning Ratio=90%2026.06 | 35.67 | 77.13 | 56.4 | |
| FastVBase Model=DriveMM, Retained Tokens=72, Pruning Ratio=90%2026.06 | 33.07 | 76.15 | 54.61 | |
| PACTBase Model=DriveMM, Retained Tokens=72, Pruning Ratio=90%2026.06 | 30.87 | 74.78 | 52.83 |