Spatial Reasoning on Ego3D-Bench
71.3Ego Distance Scoreo3-2025-04-16
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| o3-2025-04-16Model Category=Closed-source Models2025.10 | 71.3 | 59.3 | 65.6 | 93.4 | 80.1 | 53.5 | 77.7 | 83.1 | 73 | |
| 3DThinkerBase VLM=InternVL3-78B, Training Stage=S1+S22025.10 | 69.9 | 61 | 61 | 91.9 | 88.6 | 54.8 | 75.3 | 83.9 | 73.3 | |
| 3DThinkerBackbone=Qwen3-VL-30B-A3B, Training Stage=S1+S22025.10 | 69.4 | 65.1 | 39.1 | 79.3 | 75.1 | 37.6 | 76.3 | 88 | 66.2 | |
| 3DThinkerBackbone=Qwen3-VL-32B, Training Stage=S1+S22025.10 | 69.3 | 62.2 | 48.7 | 85.1 | 73.5 | 43 | 73.1 | 90.3 | 68.2 | |
| 3DThinkerBase VLM=InternVL3-14B, Training Stage=S1+S22025.10 | 63.5 | 59.9 | 41.3 | 78.3 | 80.2 | 50 | 75.1 | 84 | 66.5 | |
| 3DThinkerBackbone=Qwen3-VL-8B, Training Stage=S1+S22025.10 | 63 | 56.8 | 38.1 | 55.5 | 66.2 | 43.2 | 71.1 | 86.1 | 60 | |
| 3DThinkerBackbone=Qwen3-VL-4B, Training Stage=S1+S22025.10 | 62.3 | 56.5 | 35.5 | 61.4 | 64.1 | 40.4 | 68.3 | 84 | 59.1 | |
| 3DThinkerBase VLM=Qwen2.5-VL-32B, Training Stage=S1+S22025.10 | 62.2 | 61.9 | 54.5 | 80.2 | 86.6 | 43.7 | 69.9 | 86 | 68.1 | |
| 3DThinkerBase VLM=Qwen2.5-VL-72B, Training Stage=S1+S22025.10 | 61.1 | 59.9 | 59.7 | 93.1 | 84.9 | 43.7 | 69.8 | 87.8 | 70 | |
| 3DThinkerBackbone=Qwen3-VL-32B, Training Stage=S12025.10 | 60.4 | 51 | 48.3 | 78.6 | 67.9 | 43.2 | 68.1 | 87.3 | 63.1 | |
| 3DThinkerBackbone=Qwen3-VL-30B-A3B, Training Stage=S12025.10 | 60.1 | 54.4 | 38.3 | 72.2 | 70.3 | 38 | 71 | 80.1 | 60.6 | |
| 3DThinkerBase VLM=InternVL3-78B, Training Stage=S12025.10 | 59.8 | 53.1 | 52.2 | 80.1 | 72.5 | 53.9 | 65.1 | 78 | 64.3 | |
| gemini-2.5-proModel Category=Closed-source Models2025.10 | 58.5 | 50.2 | 61.4 | 92.9 | 75.5 | 43.5 | 72.8 | 78.6 | 66.7 | |
| 3DThinkerBase VLM=InternVL3-14B, Training Stage=S12025.10 | 56.2 | 49.1 | 37.3 | 70 | 71.8 | 51.1 | 68 | 77.7 | 60.2 | |
| doubao-seed-1.6Model Category=Closed-source Models2025.10 | 55.2 | 50.8 | 60.5 | 89 | 67.3 | 49.8 | 71.4 | 86 | 66.3 | |
| 3DThinkerBase VLM=InternVL3-38B, Training Stage=S1+S22025.10 | 54.7 | 58.1 | 49.2 | 86.9 | 80.4 | 49.1 | 79.6 | 85.9 | 68 | |
| 3DThinkerBase VLM=InternVL3-8B, Training Stage=S1+S22025.10 | 54.6 | 56.1 | 36 | 67.2 | 69.4 | 46.7 | 71 | 81.9 | 60.4 | |
| InternVL3-78BFamily=InternVL3 Family2025.10 | 54.6 | 48.4 | 50.3 | 77.7 | 70 | 44.8 | 57 | 76.6 | 59.9 | |
| 3DThinkerBackbone=Qwen3-VL-4B, Training Stage=S12025.10 | 54 | 49.9 | 35.1 | 59.3 | 60.8 | 40.8 | 64.9 | 76.1 | 55.1 | |
| 3DThinkerBase VLM=Qwen2.5-VL-7B, Training Stage=S1+S22025.10 | 54 | 52.3 | 36.5 | 52.7 | 56.6 | 38.2 | 66 | 83.1 | 54.9 | |
| 3DThinkerBackbone=Qwen3-VL-2B, Training Stage=S1+S22025.10 | 53.1 | 51.3 | 36.4 | 61.4 | 62.8 | 38.2 | 65.3 | 80.5 | 56.1 | |
| 3DThinkerBackbone=Qwen3-VL-8B, Training Stage=S12025.10 | 52.1 | 45.9 | 37.1 | 51.5 | 61.8 | 43.2 | 67.8 | 78.2 | 54.7 | |
| 3DThinkerBase VLM=Qwen2.5-VL-32B, Training Stage=S12025.10 | 52 | 51.9 | 54.8 | 80.1 | 79.4 | 44.3 | 62 | 83.1 | 63.5 | |
| gpt-4.1Model Category=Closed-source Models2025.10 | 51.7 | 36.2 | 41.8 | 82.7 | 62.6 | 44.3 | 65.7 | 70.2 | 56.9 | |
| Qwen3-VLParameter Scale=32B2025.10 | 51.5 | 40.7 | 40.9 | 76.4 | 58.2 | 40.4 | 61.5 | 81.4 | 56.4 | |
| glm-4.5vModel Category=Closed-source Models2025.10 | 49.9 | 39.6 | 48.4 | 88.8 | 73.4 | 40.4 | 57.1 | 81.9 | 59.9 | |
| 3DThinkerBase VLM=Qwen2.5-VL-72B, Training Stage=S12025.10 | 49.9 | 45.9 | 57.8 | 85.6 | 75.6 | 43.9 | 58 | 80.8 | 62.2 | |
| Qwen3-VLParameter Scale=30B-A3B2025.10 | 49.1 | 45.8 | 28.4 | 64.5 | 60 | 34.7 | 63.3 | 73.3 | 52.4 | |
| claude-sonnet-4Model Category=Closed-source Models2025.10 | 48.9 | 36.5 | 51.6 | 81.9 | 55.1 | 33.6 | 53.9 | 69.5 | 53.9 | |
| 3DThinkerBase VLM=Qwen2.5-VL-7B, Training Stage=S12025.10 | 47.9 | 44.5 | 36.5 | 51.9 | 51.3 | 39.1 | 59.1 | 73.9 | 50.5 | |
| InternVL3-14BFamily=InternVL3 Family2025.10 | 46 | 35.6 | 35.9 | 63.2 | 65.9 | 41.6 | 55.5 | 70.1 | 51.7 | |
| 3DThinkerBackbone=Qwen3-VL-2B, Training Stage=S12025.10 | 45.9 | 41.6 | 36 | 60.6 | 59.2 | 37.2 | 62 | 72.1 | 51.8 | |
| Qwen2.5-VL-32BFamily=Qwen2.5-VL Family2025.10 | 45.4 | 40.7 | 49.6 | 75.6 | 74.1 | 40.1 | 54 | 79 | 57.3 | |
| 3DThinkerBase VLM=InternVL3-38B, Training Stage=S12025.10 | 44.8 | 47 | 43.6 | 73.1 | 68.6 | 48.5 | 71.2 | 79.1 | 59.5 | |
| 3DThinkerBase VLM=InternVL3-8B, Training Stage=S12025.10 | 43.8 | 44.4 | 32.9 | 60.6 | 61.2 | 46.9 | 64.1 | 72.1 | 53.3 | |
| Qwen2.5-VL-72BFamily=Qwen2.5-VL Family2025.10 | 42.4 | 38.6 | 54.8 | 86.8 | 68.9 | 38.5 | 53.3 | 80.5 | 58 | |
| 3DThinkerBase VLM=Qwen2.5-VL-3B, Training Stage=S1+S22025.10 | 41.6 | 46 | 33.1 | 54.7 | 53.3 | 30.8 | 70.1 | 76.9 | 50.8 | |
| Qwen3-VLParameter Scale=4B2025.10 | 41.3 | 40.9 | 27.9 | 56.9 | 56.6 | 37.8 | 60.1 | 68.2 | 48.7 | |
| 3DThinker-S1+S2Model Scale=4B, Training Stage=S1 + S2, Base Model=LLaVA-OneVision-1.52025.10 | 40.2 | 39.9 | 34.2 | 51.9 | 52.3 | 30.8 | 61.8 | 73.8 | 48.1 | |
| 3DThinker-S1Model Scale=4B, Training Stage=S1, Base Model=LLaVA-OneVision-1.52025.10 | 40 | 39.1 | 33.1 | 51.1 | 52.6 | 30.9 | 58.6 | 73.8 | 47.4 | |
| LLaVA-OneVision-1.5-4BModel Scale=4B, Training Stage=Base2025.10 | 39.7 | 37.1 | 29.2 | 51.4 | 51.8 | 34.1 | 52.4 | 73.5 | 46.2 | |
| Qwen3-VLParameter Scale=8B2025.10 | 38.9 | 32.6 | 31.6 | 45.3 | 56.8 | 40.8 | 62.4 | 69.3 | 47.2 | |
| 3DThinker-S1+S2Model Scale=8B, Training Stage=S1 + S2, Base Model=LLaVA-OneVision-1.52025.10 | 36.5 | 41.5 | 37 | 46.2 | 53.3 | 32.8 | 64.9 | 77.2 | 48.7 | |
| 3DThinkerBase VLM=Qwen2.5-VL-3B, Training Stage=S12025.10 | 36.1 | 39.4 | 32.5 | 54.8 | 46.2 | 30.8 | 64 | 69.7 | 46.7 | |
| InternVL3-38BFamily=InternVL3 Family2025.10 | 35.4 | 31 | 39.4 | 66.6 | 64.9 | 38 | 61 | 77.3 | 51.7 | |
| 3DThinker-S1Model Scale=8B, Training Stage=S1, Base Model=LLaVA-OneVision-1.52025.10 | 35.1 | 39 | 36.1 | 44.9 | 53.2 | 31.9 | 61 | 73.8 | 46.9 | |
| gpt-4o-2024-11-20Model Category=Closed-source Models2025.10 | 33.2 | 26.5 | 28.1 | 78.1 | 56.7 | 36 | 60.5 | 66 | 48.1 | |
| Qwen2.5-VL-7BFamily=Qwen2.5-VL Family2025.10 | 32.7 | 31.5 | 30.5 | 45.9 | 44 | 34.5 | 43.2 | 66.5 | 41.1 | |
| LLaVA-OneVision-1.5-8BModel Scale=8B, Training Stage=Base2025.10 | 30.3 | 36.6 | 34.3 | 44.9 | 51.9 | 36.9 | 53.4 | 74.4 | 45.3 | |
| InternVL3-8BFamily=InternVL3 Family2025.10 | 25.8 | 28.7 | 29.8 | 54.1 | 54.8 | 36.1 | 49.9 | 65.2 | 43.1 | |
| Qwen3-VLParameter Scale=2B2025.10 | 24.7 | 23.9 | 30.4 | 58.7 | 56.4 | 32.8 | 57.4 | 61.1 | 43.2 | |
| Qwen2.5-VL-3BFamily=Qwen2.5-VL Family2025.10 | 21.5 | 29.4 | 28.8 | 50.3 | 41.9 | 30.9 | 54.1 | 56.1 | 39.1 |