Chinese Multi-modal Multi-task Understanding on CMMMU
42.5AccuracyGPT-4V
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-4VLLM Size=Unk2024.03 | 42.5 | |
| InternVL3.5Parameter Scale=8B2026.01 | 40 | |
| Ovis2.5Parameter Scale=9B2026.01 | 40 | |
| Qwen-VL-PlusLLM Size=Unk2024.03 | 39.5 | |
| DeepSeek-VLLLM Size=7B2024.03 | 37.9 | |
| GLM-4.6V-FlashX2026.01 | 36.3 | |
| Yi-VLLLM Size=6B2024.03 | 35.8 | |
| Ostrakon-VLParameter Scale=8B2026.01 | 33.2 | |
| Qwen3-VLParameter Scale=8B2026.01 | 33.1 | |
| Qwen2.5-VLParameter Scale=72B2026.01 | 33 | |
| DeepSeek-VLLLM=1.3B2024.03 | 27.4 | |
| CogVLMLLM Size=7B2024.03 | 24.8 | |
| EMU2-ChatLLM Size=7B2024.03 | 23.8 |