Multi-view Spatial Reasoning on MMSI (Granular Metrics)
48.77Overall ScoreSATURN
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SATURN2026.06 | 48.77 | 53.19 | 50.44 | 45.38 | 38.05 | |
| GCAData Source=Reproduced2026.06 | 41.9 | 46.74 | 44 | 36.92 | 30.81 | |
| pySpatialData Source=Paper-reported2026.06 | 37.32 | 34.87 | 42.05 | 43.1 | 36.4 | |
| Gemini-2.5-Pro2026.06 | 36.9 | 39 | 33.3 | 36.2 | 34.3 | |
| GLM-4.5V2026.06 | 33.8 | 35.6 | 29.3 | 36.9 | 30.3 | |
| Qwen3-VL-235B-ThinkingMode=Thinking2026.06 | 32.6 | 33.7 | 23.3 | 40 | 31.8 | |
| Qwen3-VL-8B-InstructMode=Instruct2026.06 | 29.4 | 32.18 | 27.33 | 31.54 | 22.22 | |
| TIGeRData Source=Paper-reported2026.06 | 27.8 | 29.1 | 26 | 27.7 | 25.8 | |
| SpaceThinker-3B2026.06 | 27.6 | 27.78 | 24.67 | 30.77 | 27.27 | |
| SpaceQwen2.5-VL-3B2026.06 | 27.1 | 28.93 | 20 | 30 | 25.76 | |
| Cosmos-Reason1-7B2026.06 | 26.9 | 28.54 | 22.67 | 28.46 | 24.75 | |
| SpaceOm2026.06 | 26.6 | 28.35 | 20.67 | 26.15 | 26.77 |