Spatial Reasoning on MMSI-Bench
97.2AccuracyHuman
Evaluation results
| Method | Links | |
|---|---|---|
| Human2026.02 | 97.2 | |
| GEMINI-3-PRO-PREVIEW2026.02 | 45.2 | |
| GPT-52026.02 | 41.8 | |
| GPT-52026.04 | 41.8 | |
| SEED-1.62026.02 | 38.3 | |
| Seed-1.62026.04 | 38.3 | |
| GEMINI-2.5-PRO2026.02 | 38 | |
| Gemini-2.5-Pro2026.04 | 38 | |
| Qwen3-VL + OursModel Size=8B2026.03 | 32.6 | |
| VST-7B-SFTEvaluation Protocol=reduced subset2026.02 | 32 | |
| VST-SFTModel Size=7B2026.03 | 32 | |
| GeoThinker QWEN2.5VL-7BActive Perception=true2026.02 | 31.7 | |
| Qwen3-VLModel Size=8B2026.03 | 31.1 | |
| Qwen3-VL-8BParameters=8B2026.05 | 31.1 | |
| BAGEL-7B-MOT2026.02 | 31 | |
| Bagel-7B-MOTSize=7B2026.04 | 31 | |
| GeoThinker QWEN3VL-8BActive Perception=true2026.02 | 30.9 | |
| GPT-4o2026.05 | 30.3 | |
| VST-3B-SFTEvaluation Protocol=reduced subset2026.02 | 30.2 | |
| VST-SFTModel Size=3B2026.03 | 30.2 | |
| GUIDE-5BSize=5B2026.04 | 30.2 | |
| VG-LLM-8B*Active Perception=false, Training Setting=S1+S22026.02 | 30 | |
| VG-LLM-8B*Size=8B, Dataset Config=S1+S22026.04 | 30 | |
| GUIDE-9BSize=9B2026.04 | 29.6 | |
| GR3D-8BParameters=8B, Backbone=Qwen-2-7B, Input Resolution=4482026.05 | 29.2 | |
| QWEN3-VL-8B-INSTRUCT2026.02 | 28.8 | |
| Qwen3-VL-8B-InstructSize=8B2026.04 | 28.8 | |
| InternVL2.5-8BParameters=8B2026.05 | 28.7 | |
| QWEN2.5-VL-3B-INSTRUCT2026.02 | 28.6 | |
| VG-LLM-8BActive Perception=false2026.02 | 28.4 | |
| VG-LLM-8BSize=8B2026.04 | 28.4 | |
| before RLBackbone=Mimo-VL-7B2025.08 | 28.2 | |
| NVILA-8BParameters=8B2026.05 | 28.1 | |
| INTERNVL3-8B2026.02 | 28 | |
| VG-LLM-4BActive Perception=false2026.02 | 28 | |
| Intern3-VL-8BModel Size=8B2026.03 | 28 | |
| InternVL3-8BSize=8B2026.04 | 28 | |
| VG-LLM-4BSize=4B2026.04 | 28 | |
| Vision-SR1Backbone=Mimo-VL-7B2025.08 | 28 | |
| VLM-3R-7BActive Perception=false2026.02 | 27.9 | |
| VLM-3R-7BSize=7B2026.04 | 27.9 | |
| Vision-R1Backbone=Mimo-VL-7B2025.08 | 27.7 | |
| Vision-SR1Backbone=Qwen2.5-VL-7B2025.08 | 27.7 | |
| Qwen3-VL + OursModel Size=2B2026.03 | 27.6 | |
| SPATIALLADDER-3B2026.02 | 27.4 | |
| SpaceR-SFT-7BModel Size=7B2026.03 | 27.4 | |
| SpatialLadder-3BSize=3B2026.04 | 27.4 | |
| Spatial-MLLM2026.03 | 27 | |
| QWEN2.5-VL-7B-INSTRUCT2026.02 | 26.8 | |
| INTERNVL3-2B2026.02 | 26.5 | |
| InternVL3-2BSize=2B2026.04 | 26.5 | |
| Qwen3-VLModel Size=2B2026.03 | 26.1 | |
| Spatial-MLLM-4BSize=4B2026.04 | 26.1 | |
| Qwen2.5-VL-7BParameters=7B2026.05 | 25.9 | |
| CAMBRIAN-S-7B2026.02 | 25.8 | |
| Cambrian-SModel Size=7B2026.03 | 25.8 | |
| SR-3D-8BParameters=8B2026.05 | 25.8 | |
| CAMBRIAN-S-3B2026.02 | 25.2 | |
| Cambrian-SModel Size=3B2026.03 | 25.2 | |
| Cambrian-S-3BSize=3B2026.04 | 25.2 | |
| Random Choice2026.02 | 25 | |
| Random Choice2026.04 | 25 | |
| before RLBackbone=Qwen2.5-VL-7B2025.08 | 24 | |
| Vision-R1Backbone=Qwen2.5-VL-7B2025.08 | 21.9 | |
| QWEN3-VL-2B-INSTRUCT2026.02 | 11.9 |