Inference Efficiency on LLaVA 7B 1.5
802.65Latency (ms)TrimTokenator-LC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TrimTokenator-LCBackbone=LLaVA-1.5-7B2025.12 | 802.65 | 15.18 | |
| DivPruneBackbone=LLaVA-1.5-7B2025.12 | 810.46 | 15.2 | |
| VisionZipBackbone=LLaVA-1.5-7B2025.12 | 813.05 | 15.2 | |
| VTWBackbone=LLaVA-1.5-7B2025.12 | 813.85 | 15.27 | |
| SparseVLMBackbone=LLaVA-1.5-7B2025.12 | 821.01 | 15.33 | |
| FastVBackbone=LLaVA-1.5-7B2025.12 | 825.91 | 15.34 | |
| CATPBackbone=LLaVA-1.5-7B2025.12 | 827.92 | 15.37 | |
| OriginalBackbone=LLaVA-1.5-7B2025.12 | 1,035.47 | 19.38 |