Multilingual Vision-Language Reasoning on MMBench Multilingual
89Accuracy (en)LaV-CoT-3B
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| LaV-CoT-3Btraining=SFT + GRPO2025.09 | 89 | 89.4 | 85.9 | 88.3 | 86.8 | 87.7 | — | |
| Gemini-2.5-flash2025.09 | 88.3 | 89 | 86.1 | 84.2 | 85.8 | 88.1 | — | |
| GPT-4o-2024-05-132025.09 | 87.6 | 88.2 | 85.5 | 85.6 | 82.9 | 86.2 | — | |
| LaV-CoT-3Btraining=SFT2025.09 | 85.7 | 84.4 | 84.3 | 83.9 | 85.2 | 83.1 | — | |
| InternVL3-8B2025.09 | 85.5 | 85.6 | 83.2 | 79.2 | 75.9 | 82.6 | — | |
| Qwen2.5-VL-7B2025.09 | 85.3 | 85.8 | 83 | 80.2 | 75.7 | 82.9 | — | |
| InternVL3.5-8B2025.09 | 84.5 | 85.7 | 80.9 | 82.8 | 75.8 | 82.3 | — | |
| Eagle2-9B2025.09 | 83.8 | 83.7 | 81.5 | 78 | 75.1 | 81.1 | — | |
| Qwen2-VL-7B2025.09 | 81.8 | 81.6 | 79.1 | 75.6 | 74.5 | 79.3 | — | |
| InternVL3-2B2025.09 | 81.3 | 77.8 | 75.9 | 66.4 | 59.5 | 70.7 | — | |
| Qwen2.5-VL-3B2025.09 | 81.1 | 79.7 | 75.8 | 69.3 | 62.4 | 72.6 | — | |
| InternVL3.5-2B2025.09 | 79 | 76.5 | 74.3 | 64.4 | 63.1 | 72.2 | — | |
| PARROT-7B2025.09 | 78 | 77.1 | 76.7 | 75.9 | 74 | 77.7 | — | |
| LLaVA-OneVision-7B2025.09 | 76.7 | 75.3 | 73.4 | 70.4 | 64.8 | 73.1 | — | |
| Qwen2-VL-2B2025.09 | 72.1 | 71.1 | 69.9 | 61.1 | 54.4 | 69.3 | — | |
| DeepSeek-VL-7B2025.09 | 70.7 | 64 | 62.6 | 48 | 47.9 | 65.5 | — | |
| GLM-4v-9B2025.09 | 67.9 | 61.3 | 60 | 46.1 | 45.7 | 63 | — | |
| Monkey-9.8B2025.09 | 58 | 53.5 | 49.5 | 31 | 31.3 | 45.1 | — |