Video spatial & 4D reasoning on VSI-Bench-U 2025
58.8AccuracyNo-tool
Evaluation Results
| Method | Links | |
|---|---|---|
| No-toolModel=Qwen3.5-397B-A17B2026.06 | 58.8 | |
| SpatialClawModel=Qwen3.6-27B2026.06 | 57.9 | |
| SpatialClawModel=Qwen3.5-397B-A17B2026.06 | 56.6 | |
| No-toolModel=Qwen3.5-122B-A10B2026.06 | 55.7 | |
| No-toolModel=Qwen3.6-27B2026.06 | 55.4 | |
| No-toolModel=Qwen3.6-35B-A3B2026.06 | 54.2 | |
| SpatialClawModel=Qwen3.6-35B-A3B2026.06 | 51.7 | |
| SpatialClawModel=Qwen3.5-122B-A10B2026.06 | 50.6 | |
| SpatialClawModel=Gemma4-31B2026.06 | 48.5 | |
| No-toolModel=Gemma4-31B2026.06 | 48 | |
| SpatialClawModel=Gemma4-26B-A4B2026.06 | 38.5 | |
| No-toolModel=Gemma4-26B-A4B2026.06 | 30.1 |