Spatial Reasoning on Tangram one-piece
44.3Position IoUGemini-2.5-pro
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Gemini-2.5-pro2026.02 | 44.3 | 43.4 | 43.2 | 41.7 | |
| GPT-4o mini-8BParameters=8B2026.02 | 42.7 | 42.9 | 39.3 | 41.3 | |
| LLaMA Maverick 17BParameters=17B2026.02 | 42.4 | 42.7 | 37.1 | 37.7 | |
| Claude-Sonnet-42026.02 | 41.9 | 39.4 | 37.2 | 39.5 | |
| Qwen-72BParameters=72B2026.02 | 41.5 | 43.2 | 42.5 | 40.8 | |
| Qwen-3BParameters=3B2026.02 | 23.6 | 41.4 | 36.9 | 21.9 |