Visual Spatial Reasoning on What's Up (Split B)
100AccuracyHuman Estimate
Evaluation Results
| Method | Links | |
|---|---|---|
| Human Estimate2025.04 | 100 | |
| SpaRE-7BModel size=7B2025.04 | 100 | |
| GPT-4o2025.04 | 100 | |
| Qwen2VL-7BModel size=7B2025.04 | 99.3 | |
| LLaVA-NeXT-8BModel size=8B2025.04 | 95.6 | |
| SpaRE-2BModel size=2B2025.04 | 95.1 | |
| InternVL2-8BModel size=8B2025.04 | 94.4 | |
| GPT-4o-mini2025.04 | 90 | |
| InternVL2-2BModel size=2B2025.04 | 84.7 | |
| Qwen2VL-2BModel size=2B2025.04 | 79.1 | |
| SpaceLLaVa2025.04 | 75.6 | |
| InternVL3.5 8BPrompting Strategy=Task-Aware2026.01 | 46.3 | |
| BLIP 14M2026.01 | 38.2 | |
| MiniCPM-V 4.5Prompting Strategy=Task-Aware2026.01 | 36.3 | |
| XVLM 16M2026.01 | 33.1 | |
| CLIP-ViT-B/322026.01 | 32.4 | |
| MiniCPM-V 4.5Prompting Strategy=General2026.01 | 32.4 | |
| InternVL3.5 8BPrompting Strategy=General2026.01 | 27.9 | |
| Random2025.04 | 25 | |
| Random2026.01 | 25 |