Multimodal Reasoning on MMBench (Accuracy and Token Count)
82.1AccuracyLaRe
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LaReBase VLM=Qwen3-VL-4B-Instruct2025.11 | 82.1 | 49 | |
| Standard CoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 81.7 | 165.2 | |
| CCoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 80.5 | 125.4 | |
| LVR*Base VLM=Qwen3-VL-4B-Instruct2025.11 | 79 | 80.7 | |
| CoF*Base VLM=Qwen3-VL-4B-Instruct2025.11 | 78.6 | 190.8 | |
| ICoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 76.9 | 162.5 | |
| MM-CoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 75.9 | 170.6 | |
| MCOUTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 73.8 | 72.3 | |
| CoF*Base VLM=Qwen3-VL-2B-Instruct2025.11 | 73.2 | 162.6 | |
| MM-CoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 72.7 | 158.3 | |
| LaReBase VLM=Qwen3-VL-2B-Instruct2025.11 | 71.9 | 34.6 | |
| ICoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 71.5 | 133.4 | |
| Standard CoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 71.4 | 167.4 | |
| CCoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 69.9 | 115.7 | |
| LVR*Base VLM=Qwen3-VL-2B-Instruct2025.11 | 67.5 | 75 | |
| MCOUTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 62.6 | 65.3 |