Multimodal Spatial Reasoning on BLINK
84.3Average AccuracyMulti-SpatialMLLM
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Multi-SpatialMLLMevaluation_protocol=zero-shot2025.05 | 84.3 | 89.5 | 79.8 | 94.7 | 74.8 | |
| Gemini-2.0evaluation_protocol=zero-shot2025.05 | 75.7 | 88.4 | 83.9 | 42.9 | 83.9 | |
| GPT-4oevaluation_protocol=zero-shot2025.05 | 73.8 | 84.9 | 71 | 53.4 | 81.8 | |
| Claude-3.5evaluation_protocol=zero-shot2025.05 | 67.7 | 74.4 | 63.7 | 54.1 | 75.5 | |
| InternVL-26Bevaluation_protocol=zero-shot2025.05 | 61.4 | 47.1 | 78.2 | 44.4 | 79.7 | |
| InternVL-8Bevaluation_protocol=zero-shot2025.05 | 57.9 | 39 | 71.8 | 49.6 | 76.2 | |
| SpatialRGPTevaluation_protocol=zero-shot2025.05 | 52.5 | 34.3 | 61.3 | 44.4 | 74.1 |