Perspective-dependent reasoning on ViewSpatial-Bench (accuracy (%))
63AccuracyGAMSI_S1+S2
Evaluation Results
| Method | Links | |
|---|---|---|
| GAMSI_S1+S2SI Dataset=950K2026.05 | 63 | |
| GAMSI_S1SI Dataset=800K2026.05 | 59.6 | |
| SenseNova-SI-InternVL3-8BSI Dataset=800K2026.05 | 55.1 | |
| VST-3B-SFTSI Dataset=4M2026.05 | 52.9 | |
| SenseNova-SI-Qwen3-VL-8BSI Dataset=8M2026.05 | 51.2 | |
| VST-7B-SFTSI Dataset=4M2026.05 | 50.7 | |
| SpatialLadder-3BSI Dataset=260K2026.05 | 44.8 | |
| Qwen3-VL-8B-Instruct2026.05 | 42.2 | |
| Bagel-7B-MoT2026.05 | 41.3 | |
| Cambrian-S-7BSI Dataset=590K2026.05 | 41.3 | |
| Cambrian-S-3BSI Dataset=590K2026.05 | 40.9 | |
| InternVL3-8B2026.05 | 38.6 | |
| SpaceR-7BSI Dataset=151K2026.05 | 38.4 | |
| Qwen2.5-VL-7B-Instruct2026.05 | 37.4 | |
| Qwen3-VL-2B-Instruct2026.05 | 36.4 | |
| Qwen2.5-VL-3B-Instruct2026.05 | 36.2 | |
| ViLaSR-7BSI Dataset=81K2026.05 | 35.7 | |
| Spatial-MLLM-4BSI Dataset=135K2026.05 | 34.6 | |
| InternVL3-2B2026.05 | 32.7 |