Multimodal Reasoning on GQA (test)
67.4AccuracyOurs
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OursModel=Qwen2-VL 7B2026.04 | 67.4 | 9.2 | 0.82 | |
| IVT-LRModel=Qwen2-VL 7B2026.04 | 65.8 | 10.1 | 0.68 | |
| Chain-of-FocusModel=Qwen2-VL 7B2026.04 | 61.8 | 128.6 | 3.01 | |
| CCoTModel=Qwen2-VL 7B2026.04 | 51.2 | 76.4 | 7.21 | |
| SCAFFOLDModel=Qwen2-VL 7B2026.04 | 48.7 | 72.8 | 6.72 | |
| OursModel=Chameleon 7B2026.04 | 39.4 | 9.2 | 1.21 | |
| IVT-LRModel=Chameleon 7B2026.04 | 38.1 | 10.1 | 0.98 | |
| Chain-of-FocusModel=Chameleon 7B2026.04 | 34.6 | 360.4 | 2.98 | |
| CCoTModel=Chameleon 7B2026.04 | 33.1 | 150.6 | 5.31 | |
| SCAFFOLDModel=Chameleon 7B2026.04 | 32.8 | 156 | 4.17 |