Multimodal Reasoning on MMVP (Accuracy, Token Count)
57.6AccuracyLaRe
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LaReBase VLM=Qwen3-VL-4B-Instruct2025.11 | 57.6 | 32.3 | |
| LVR*Base VLM=Qwen3-VL-4B-Instruct2025.11 | 56.2 | 72.9 | |
| LaReBase VLM=Qwen3-VL-2B-Instruct2025.11 | 53.3 | 27.2 | |
| CoF*Base VLM=Qwen3-VL-4B-Instruct2025.11 | 51.8 | 145.8 | |
| CoF*Base VLM=Qwen3-VL-2B-Instruct2025.11 | 50.1 | 137.3 | |
| ICoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 46 | 122.7 | |
| MCOUTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 45.7 | 69.2 | |
| LVR*Base VLM=Qwen3-VL-2B-Instruct2025.11 | 43.9 | 72.4 | |
| ICoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 40.3 | 108.7 | |
| CCoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 39.3 | 140.9 | |
| CCoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 37.5 | 128.6 | |
| MCOUTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 37.3 | 67.9 | |
| MM-CoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 27.5 | 113 | |
| MM-CoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 24 | 107.7 | |
| Standard CoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 23.7 | 146.2 | |
| Standard CoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 18.4 | 126.5 |