Spatial Reasoning on CV-Bench 1k random samples
69.2Count ScoreSpatiO
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SpatiO2026.04 | 69.2 | 91.8 | 95.7 | 91.2 | 86.9 | |
| Qwen-3.0-VL-4BLoRA training=true2026.04 | 65.9 | 88.7 | 94 | 89.1 | 84.4 | |
| Gpt-5.22026.04 | 65.4 | 95.7 | 90.4 | 87.2 | 83.5 | |
| Qwen-3.0-VL-4BLoRA training=false2026.04 | 65 | 87.7 | 93.8 | 83.5 | 81.3 | |
| SpatialReasonerLoRA training=true2026.04 | 61.5 | 86.9 | 82.1 | 83.1 | 77.4 | |
| Llava-4DLoRA training=true2026.04 | 61.1 | 61.6 | 82.3 | 70.8 | 68.3 | |
| Claude opus 4.62026.04 | 59.2 | 94.8 | 88.8 | 86.8 | 82.4 | |
| Sa2VALoRA training=false2026.04 | 59 | 78 | 80.6 | 66.1 | 70.2 | |
| Sa2VALoRA training=true2026.04 | 59 | 78 | 80.7 | 66.2 | 70.2 | |
| SpatialReasonerLoRA training=false2026.04 | 58.9 | 85.6 | 83.4 | 82.3 | 77.6 | |
| SpatialRGPTLoRA training=false2026.04 | 51 | 84.4 | 55.3 | 56.5 | 61.4 | |
| SpatialRGPTLoRA training=true2026.04 | 48.3 | 81 | 59.6 | 57.4 | 61 | |
| Llava-4DLoRA training=false2026.04 | 40.8 | 61.3 | 51.3 | 54.9 | 51.3 |