Image Captioning on COCO 17
107CIDErCASP_AQLM
Evaluation Results
| Method | Links | |
|---|---|---|
| CASP_AQLMBit=22025.03 | 107 | |
| OriginalBit=162025.03 | 106.2 | |
| QuIP#Bit=22025.03 | 103.8 | |
| CASP_QuIP#Bit=22025.03 | 102.8 | |
| CASP_GPTQBit=2.22025.03 | 100.8 | |
| AQLMBit=22025.03 | 87.1 | |
| GPTQBit=2.22025.03 | 62.4 | |
| Vanilla (LLaVA-1.5-7B)Retained Tokens=576, Reduction Rate=0.0%2026.01 | 1.102 | |
| PyramidDropRetained Tokens=192, Reduction Rate=66.7%2026.01 | 1.091 | |
| SparseVLMRetained Tokens=192, Reduction Rate=66.7%2026.01 | 1.087 | |
| FastVRetained Tokens=192, Reduction Rate=66.7%2026.01 | 1.082 | |
| ViTCoPRetained Tokens=192, Reduction Rate=66.7%2026.01 | 1.078 | |
| VisionZipRetained Tokens=192, Reduction Rate=66.7%2026.01 | 1.07 | |
| ViTCoPRetained Tokens=128, Reduction Rate=77.8%2026.01 | 1.064 | |
| FastVRetained Tokens=128, Reduction Rate=77.8%2026.01 | 1.044 | |
| PyramidDropRetained Tokens=128, Reduction Rate=77.8%2026.01 | 1.039 | |
| VisionZipRetained Tokens=128, Reduction Rate=77.8%2026.01 | 1.037 | |
| ViTCoPRetained Tokens=64, Reduction Rate=88.9%2026.01 | 1.032 | |
| VisionZipRetained Tokens=64, Reduction Rate=88.9%2026.01 | 0.948 | |
| SparseVLMRetained Tokens=128, Reduction Rate=77.8%2026.01 | 0.94 | |
| FastVRetained Tokens=64, Reduction Rate=88.9%2026.01 | 0.815 | |
| SparseVLMRetained Tokens=64, Reduction Rate=88.9%2026.01 | 0.731 | |
| PyramidDropRetained Tokens=64, Reduction Rate=88.9%2026.01 | 0.648 |