Spatial Reasoning on SPBench SI
79.1AccuracyGAMSI_S1+S2
Evaluation Results
| Method | Links | |
|---|---|---|
| GAMSI_S1+S2SI Dataset=950K2026.05 | 79.1 | |
| SpatialLadder-3BSI Dataset=260K2026.05 | 70 | |
| OneCanvasEvaluation Protocol=Zero-shot evaluation2026.06 | 62.8 | |
| SenseNova-SI-InternVL3-8BSI Dataset=800K2026.05 | 62.5 | |
| GAMSI_S1SI Dataset=800K2026.05 | 60.6 | |
| SenseNova-SI-Qwen3-VL-8BSI Dataset=8M2026.05 | 60.4 | |
| SpaceMindEvaluation Protocol=Zero-shot evaluation2026.06 | 59.7 | |
| SpatialLadder-3BBase Model=Qwen2.5-VL-3B2026.05 | 58.6 | |
| Cambrian-S-3BSI Dataset=590K2026.05 | 57.7 | |
| Cambrian-S-7BSI Dataset=590K2026.05 | 54.6 | |
| 3ViewSense-4B-rl-slackTraining=RL, Reward=Slack2026.03 | 54.2 | |
| VST-3B-SFTSI Dataset=4M2026.05 | 53.7 | |
| 3ViewSense-4B-sftTraining=SFT2026.03 | 52.6 | |
| VST-7B-SFTSI Dataset=4M2026.05 | 52.2 | |
| 3ViewSense-4B-rl-strictTraining=RL, Reward=Strict2026.03 | 52 | |
| Qwen3-VL-8B-Instruct2026.05 | 51.7 | |
| Qwen3-VL-2B-Instruct2026.05 | 51.6 | |
| CaMo-3BBase Model=Qwen2.5-VL-3B2026.05 | 51.5 | |
| InternVL3-8B2026.05 | 51.5 | |
| VILASR-7BEvaluation Protocol=Zero-shot evaluation2026.06 | 50.2 | |
| Gemini-2.0-FlashModel Type=Proprietary2026.05 | 49 | |
| ViLaSR-7BSI Dataset=81K2026.05 | 48.9 | |
| SpaceR-7BEvaluation Protocol=Zero-shot evaluation2026.06 | 48.6 | |
| SpaceR-7BSI Dataset=151K2026.05 | 47.9 | |
| Qwen2.5-VL-7B-Instruct2026.05 | 47.4 | |
| Spatial-MLLM-4BSI Dataset=135K2026.05 | 45.3 | |
| Video-R1Evaluation Protocol=Zero-shot evaluation2026.06 | 44.9 | |
| Spatial-MLLM-4BEvaluation Protocol=Zero-shot evaluation2026.06 | 43.7 | |
| Qwen2.5-VL-3BModel=Qwen2.5-VL-3B2026.03 | 43.5 | |
| InternVL3-2B2026.05 | 41.2 | |
| Qwen2.5-VL-3B-Instruct2026.05 | 40.9 | |
| SpatialLadder-3BModel=SpatialLadder-3B2026.03 | 40.8 | |
| Spatial-MLLM-4BBase Model=Qwen2.5-VL-3B2026.05 | 38.1 | |
| Bagel-7B-MoT2026.05 | 38.1 | |
| Qwen3-VL-8B-InstructModel=Qwen3-VL-8B-Instruct2026.03 | 37.6 | |
| VILASR-7BBase Model=Qwen2.5-VL-7B2026.05 | 36.6 | |
| Qwen2.5-VL-7BBase Model=Qwen2.5-VL-7B2026.05 | 36.3 | |
| SpaceR-7BBase Model=Qwen2.5-VL-7B2026.05 | 35.7 | |
| InternVL-2.5-4BModel Type=Open-Source2026.05 | 31.8 | |
| SpaceQwen2.5-VL-3B-InstructModel=SpaceQwen2.5-VL-3B-Instruct2026.03 | 29.7 | |
| InternVL-2.5-8BModel Type=Open-Source2026.05 | 28.3 | |
| InternVL3.5-4BModel=InternVL3.5-4B2026.03 | 28.1 | |
| Video-R1Base Model=Qwen2.5-VL-7B2026.05 | 27.7 | |
| Qwen3-VL-4B-InstructModel=Qwen3-VL-4B-Instruct2026.03 | 27.1 | |
| Kimi-VL-A3BModel Type=Open-Source2026.05 | 25.7 | |
| LLaVA-OneVision-7BModel Type=Open-Source2026.05 | 25.4 | |
| GPT-4oModel Type=Proprietary2026.05 | 24.5 | |
| Qwen2.5-VL-3BBase Model=Qwen2.5-VL-3B2026.05 | 24.3 |