Spatial Reasoning on 3D FORCE SAG 1.0 (test)
94.4SAG ScoreSATURN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SATURNOracle 3D=true2026.06 | 94.4 | 95.94 | |
| SATURNBackbone=Qwen3-VL-8B-Instruct2026.06 | 87.57 | 88.85 | |
| Gemini-3.1-Pro2026.06 | 75.3 | 57.41 | |
| Qwen3.5-9BParameters=9B2026.06 | 74.78 | 67.08 | |
| Qwen3-VL-235BParameters=235B2026.06 | 73.22 | 26.87 | |
| GPT-5.12026.06 | 71.39 | 27.46 | |
| Qwen3-VL-8BParameters=8B2026.06 | 64.17 | 24.15 | |
| Qwen3.5-35BParameters=35B2026.06 | 58.52 | 64.33 | |
| InternVL3.5-38BParameters=38B2026.06 | 58.17 | 33.35 | |
| Molmo2-8BParameters=8B2026.06 | 54.61 | 20.32 | |
| InternVL3.5-8BParameters=8B2026.06 | 53.3 | 27.33 | |
| GCAtype=tool-augmented2026.06 | 51 | 24.24 | |
| Random choice2026.06 | 50.17 | 18.8 | |
| SpaceOm-3BParameters=3B2026.06 | 49.39 | 17.79 | |
| SpaceThinker-3BParameters=3B2026.06 | 47.91 | 17.45 | |
| SpaceQwen-3BParameters=3B2026.06 | 37.74 | 13.96 | |
| Cosmos-Reason1-7BParameters=7B2026.06 | 36.87 | 13.25 |