Causal Spatial Reasoning on CausalSpatial Level 2
78.57Collision AnticipationHuman
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Human2026.01 | 78.57 | 97.22 | 97.5 | 62.5 | 83.54 | — | |
| Gemini2.5 Flash2026.01 | 42.31 | 58.33 | 25.93 | 55.47 | 46.53 | 1.79 | |
| GPT52026.01 | 40.94 | 50 | 62.96 | 64.96 | 54.52 | 1.08 | |
| GPT5-mini2026.01 | 35.43 | 47.22 | 50 | 67.88 | 51.41 | 3.94 | |
| Qwen3-VL-30B-A3B-T.Scale=30B-A3B, Reasoning_Trained=true2026.01 | 35.09 | 63.89 | 46.3 | 48.18 | 44.79 | 0.36 | |
| Gemini2.5 Pro2026.01 | 30.65 | 52.78 | 48.15 | 64.96 | 48.85 | 0.36 | |
| Qwen2.5-VL-7B-I.Scale=7B2026.01 | 30.4 | 41.67 | 24.07 | 38.69 | 33.79 | 5.73 | |
| Qwen3-VL-30B-A3B-I.Scale=30B-A3B2026.01 | 30.16 | 52.78 | 31.48 | 52.55 | 41.32 | 1.08 | |
| Sonnet 3.72026.01 | 29.41 | 48.61 | 34.26 | 60.58 | 44.16 | 17.29 | |
| Spatial-MLLM2026.01 | 24.41 | 44.44 | 12.96 | 24.82 | 24.56 | 0.72 | |
| Qwen3-VL-4B-I.Scale=4B2026.01 | 24.41 | 44.44 | 16.67 | 62.77 | 40.11 | 10.39 | |
| Qwen3-VL-8B-T.Scale=8B, Reasoning_Trained=true2026.01 | 23.75 | 44.44 | 22.22 | 34.31 | 29.71 | 1.79 | |
| Qwen3-VL-8B-I.Scale=8B2026.01 | 22.76 | 36.11 | 29.63 | 50.36 | 35.85 | 7.89 | |
| Qwen3-VL-2B-I.Scale=2B2026.01 | 19.35 | 38.89 | 22.22 | 37.96 | 28.98 | 1.79 | |
| Spatial-VLM2026.01 | 15.75 | 41.67 | 24.07 | 37.96 | 28.25 | 7.17 | |
| LLaVAOneVision1.5-8B-I.Scale=8B2026.01 | 15.38 | 41.67 | 20.37 | 38.69 | 27.84 | 23.3 | |
| Random2026.01 | 8.54 | 33.33 | 20.37 | 32.14 | 24.5 | — |