Image Captioning on NoCaps 4,500 (test)
122.1CIDErQwen2.5 VL-7B + PHM
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen2.5 VL-7B + PHMBase Model=Qwen2.5 VL-7B, Compression=PHM, Parameters=~5.38B2025.12 | 122.1 | 89.2 | 47.1 | 30.4 | 62.3 | |
| Qwen2.5 VL-7BParameters=7B2025.12 | 121.4 | — | — | — | — | |
| InstructBLIP-7BParameters=7B2025.12 | 118.2 | 88.2 | 47.2 | 30.3 | 62 | |
| InstructBLIP-7B + PHMBase Model=InstructBLIP-7B, Compression=PHM, Parameters=~5.45B2025.12 | 116.8 | 87.6 | 46.5 | 29.9 | 61.5 | |
| InstructBLIP-13BParameters=13B2025.12 | 116.3 | 88.3 | 46.7 | 29.8 | 61.4 | |
| LLaVA-1.5-13BParameters=13B2025.12 | 109.2 | 84.2 | 42.4 | 30.6 | 60.3 | |
| LLaVA-1.5-7BParameters=7B2025.12 | 105.5 | 82.6 | 40.2 | 30.3 | 59.4 | |
| LLaVA-1.5-7B + PHMBase Model=LLaVA-1.5-7B, Compression=PHM, Parameters=~5.15B2025.12 | 104 | 81.8 | 39.6 | 29.8 | 58.7 | |
| MiniCPM-V-3BParameters=3B2025.12 | 89.9 | 79.1 | 33.2 | 29.7 | 55.8 | |
| LLaVA-1.6-7BParameters=7B2025.12 | 88.4 | 73.8 | 34.8 | 25.9 | 54.6 | |
| LLaVA-1.6-13BParameters=13B2025.12 | 88.1 | 68.7 | 34 | 25.4 | 54.9 | |
| DeCap2025.12 | 42.7 | — | — | — | — |