Visual Spatial Reasoning on MMSI-Bench
36.6Cam-Cam AccuracyGPT-4.1
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GPT-4.1Model=GPT-4.12026.02 | 36.6 | 26.6 | 27.2 | 29.1 | 36.5 | 27.7 | 37.5 | 24.2 | 36.5 | 32.9 | 28.8 | 30.9 | |
| GPT-4oModel=GPT-4o2026.02 | 34.4 | 24.5 | 23.5 | 19.8 | 37.6 | 27.7 | 32.8 | 31.8 | 35.1 | 36.8 | 30.8 | 30.3 | |
| AVICBackbone=GPT-4.12026.02 | 32.2 | 24.4 | 27.1 | 26.7 | 44.7 | 32.5 | 51.5 | 37.8 | 33.7 | 36.8 | 32.3 | 33.8 | |
| AVICBackbone=GPT-4o2026.02 | 31.1 | 29.9 | 27.8 | 25 | 36.5 | 36.5 | 51.2 | 39.7 | 31.6 | 32.3 | 28.2 | 32.3 |