Spatial Alignment on spatial alignment benchmark
80.9Mean IoUQwen2.5-7B (3D-Layout-R1)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-7B (3D-Layout-R1)Training/Evaluation Protocol=Ours2026.03 | 80.9 | 80.3 | 0.176 | |
| Qwen2.5-VL-7B (3D-Layout-R1)Training/Evaluation Protocol=Ours2026.03 | 79.8 | 79.2 | 0.181 | |
| Qwen3-8B (3D-Layout-R1)Training/Evaluation Protocol=Ours2026.03 | 79 | 78.4 | 0.19 | |
| Qwen3-VL-235B-Instruct2026.03 | 78.6 | 78.7 | 0.177 | |
| Qwen2.5-VL-7BTraining/Evaluation Protocol=CoT-SFT2026.03 | 78.5 | 77.9 | 0.193 | |
| Qwen2.5-7BTraining/Evaluation Protocol=CoT-SFT2026.03 | 76.4 | 75.8 | 0.189 | |
| Gemini 2.5 Pro2026.03 | 75.9 | 77.2 | 0.156 | |
| Qwen2.5-7BTraining/Evaluation Protocol=Ans-SFT2026.03 | 73.3 | 72.9 | 0.201 | |
| Qwen2.5-VL-7BTraining/Evaluation Protocol=Ans-SFT2026.03 | 72.8 | 72.5 | 0.201 | |
| Qwen3-8BTraining/Evaluation Protocol=CoT-SFT2026.03 | 70.7 | 70 | 0.227 | |
| Qwen3-8BTraining/Evaluation Protocol=Ans-SFT2026.03 | 64.9 | 63.6 | 0.248 | |
| Qwen3-VL-235B-Thinking2026.03 | 62.3 | 62.4 | 0.189 | |
| Gemini 2.5 Flash2026.03 | 61.3 | 61.1 | 0.231 | |
| Deepseek-R12026.03 | 60.2 | 60.7 | 0.184 | |
| Deepseek-R1-05282026.03 | 59.9 | 60.6 | 0.252 | |
| Qwen3-235B-Instruct2026.03 | 59.2 | 58.7 | 0.243 | |
| Deepseek-V32026.03 | 55.3 | 55.9 | 0.278 | |
| Qwen3-235B-Thinking2026.03 | 44.5 | 42.3 | 0.32 |