Causal Spatial Reasoning on CausalSpatial Level 1
86.84Collision AnticipationHuman
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Human2026.01 | 86.84 | 80 | 92.5 | 82.5 | 85.44 | — | |
| Sonnet 3.72026.01 | 55 | 49.21 | 45.56 | 46.35 | 50.12 | 5.04 | |
| Qwen3-VL-30B-A3B-T.Scale=30B-A3B, Reasoning_Trained=true2026.01 | 52.99 | 58.73 | 51.11 | 54.17 | 53.49 | 0 | |
| GPT52026.01 | 49.65 | 61.9 | 59.26 | 54.17 | 54 | 1.47 | |
| Qwen3-VL-4B-I.Scale=4B2026.01 | 47.55 | 50.79 | 48.15 | 44.79 | 47.19 | 6.92 | |
| Qwen3-VL-2B-I.Scale=2B2026.01 | 47.43 | 47.62 | 49.63 | 43.23 | 46.69 | 0.88 | |
| Qwen3-VL-30B-A3B-I.Scale=30B-A3B2026.01 | 47.02 | 55.56 | 48.15 | 47.4 | 48.15 | 1.47 | |
| Gemini2.5 Flash2026.01 | 46.21 | 50.79 | 48.89 | 56.25 | 50.02 | 0.15 | |
| Gemini2.5 Pro2026.01 | 43.62 | 63.49 | 60 | 53.12 | 51.44 | 0 | |
| LLaVAOneVision1.5-8B-I.Scale=8B2026.01 | 43.01 | 42.86 | 38.52 | 29.17 | 38.17 | 15.91 | |
| Qwen3-VL-8B-T.Scale=8B, Reasoning_Trained=true2026.01 | 42.74 | 63.49 | 48.15 | 55.73 | 49.43 | 1.03 | |
| Spatial-MLLM2026.01 | 41.96 | 55.56 | 48.15 | 54.17 | 47.93 | 1.91 | |
| Qwen2.5-VL-7B-I.Scale=7B2026.01 | 41.75 | 42.86 | 42.22 | 44.27 | 42.66 | 2.21 | |
| Qwen3-VL-8B-I.Scale=8B2026.01 | 41.4 | 66.67 | 51.11 | 47.92 | 47.55 | 2.95 | |
| GPT5-mini2026.01 | 37.76 | 42.86 | 48.15 | 44.79 | 42.31 | 6.19 | |
| Random2026.01 | 23.61 | 31.75 | 32.35 | 38.54 | 30.34 | — | |
| Spatial-VLM2026.01 | 20.98 | 47.62 | 48.15 | 46.88 | 36.24 | 3.83 |