Visual Number Reasoning on VisNum
47.94AccuracyActive-Zero
Evaluation Results
| Method | Links | |
|---|---|---|
| Active-ZeroBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 47.94 | |
| VisPlayBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 45.22 | |
| EvoLMMBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 45.01 | |
| Base ModelBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 44.85 | |
| VisionZero-ChartBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 44.54 | |
| VisionZero-RealWorldBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 44.17 | |
| VisionZero-CLEVRBackbone=Qwen2.5-VL-7B-Instruct2026.02 | 43.13 | |
| DUELRole=Solver2026.05 | 42.7 | |
| Vision-Zero(CLEVR)2026.05 | 42.3 | |
| OpenVLThinker-7B2026.05 | 42.2 | |
| EvoLMM2026.05 | 41.9 | |
| DUELRole=Challenger2026.05 | 41.9 | |
| VLAA-Thinker-7B2026.05 | 41.8 | |
| Qwen2.5-VL-7B2026.05 | 41.7 | |
| MM-UPT2026.05 | 41.6 | |
| Active-ZeroBackbone=Qwen2.5-VL-3B-Instruct2026.02 | 41.51 | |
| VisPlayBackbone=Qwen2.5-VL-3B-Instruct2026.02 | 37.32 | |
| Base ModelBackbone=Qwen2.5-VL-3B-Instruct2026.02 | 34.71 |