Image & Text to Text (IT2T) on Oogiri-GO (test)
41.83T1Qwen-VL+CLoT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen-VL+CLoTSize=7B2023.12 | 41.8 | 38.7 | 21.6 | 68.5 | 42.7 | |
| Qwen-VL+AITSize=7B2023.12 | 39.7 | 38.9 | — | 67.3 | 40.4 | |
| Qwen-VLSize=7B2023.12 | 30.2 | 26 | 10.4 | 67.7 | 33.6 | |
| mPLUG-Owl MultilingualSize=7B2023.12 | 28.1 | 26 | 10.5 | 64.4 | 32.2 | |
| VisualGLM-6BSize=6B2023.12 | 24.1 | 22.5 | 9.7 | 67.4 | 30.9 | |
| GPT4v2023.12 | 19.3 | 14.9 | 3.2 | 56.7 | 23.5 | |
| LLaVA-1.5Size=13B2023.12 | 13.2 | 13.7 | 13.9 | 68.1 | 27.2 | |
| MiniGPT-v2Size=7B2023.12 | 6.1 | 3.4 | 4 | 60.7 | 18.6 |