Multimodal Reasoning on V* (Acc., # Tokens)
87AccuracyLaRe
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LaReBase VLM=Qwen3-VL-4B-Instruct2025.11 | 87 | 93.5 | |
| CoF*Base VLM=Qwen3-VL-4B-Instruct2025.11 | 83.7 | 271.1 | |
| LVR*Base VLM=Qwen3-VL-4B-Instruct2025.11 | 83.7 | 110.9 | |
| MCOUTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 82.5 | 104.8 | |
| CCoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 81.5 | 203.4 | |
| LaReBase VLM=Qwen3-VL-2B-Instruct2025.11 | 81.3 | 87.2 | |
| ICoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 80.4 | 195 | |
| Standard CoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 80 | 183.4 | |
| LVR*Base VLM=Qwen3-VL-2B-Instruct2025.11 | 77.4 | 112.7 | |
| CoF*Base VLM=Qwen3-VL-2B-Instruct2025.11 | 77.3 | 255.9 | |
| MCOUTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 76.5 | 99.2 | |
| ICoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 76.3 | 199.2 | |
| CCoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 75.2 | 221.4 | |
| Standard CoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 75 | 181.2 | |
| MM-CoTBase VLM=Qwen3-VL-4B-Instruct2025.11 | 74.5 | 187.2 | |
| MM-CoTBase VLM=Qwen3-VL-2B-Instruct2025.11 | 66.5 | 192.3 |