Fine-grained Video Spatial Reasoning on VSI-Bench
60.6Avg ScoreVST-7B-SFT
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| VST-7B-SFTData Scale=4.1M, Training Phases=1 (S)2026.01 | 60.6 | 72 | 44.4 | 74.3 | 68.3 | 59.7 | 55.8 | 44.9 | 65.2 | |
| SmoothOp-7BData Scale=50K, Training Phases=1 (R)2026.01 | 60 | 71.1 | 41.4 | 72.2 | 70.3 | 61.7 | 55.1 | 42.8 | 65.4 | |
| VST-3B-SFTData Scale=4.1M, Training Phases=1 (S)2026.01 | 57.9 | 69.3 | 45.4 | 71.8 | 62.4 | 59 | 46 | 38.7 | 70.2 | |
| SmoothOp-3BData Scale=50K, Training Phases=1 (R)2026.01 | 53.9 | 70.1 | 42.8 | 69.9 | 38.3 | 60.4 | 45.4 | 38.7 | 65.5 | |
| InternVL-Spatial-8BData Scale=291K, Training Phases=1 (S)2026.01 | 52.3 | 68.7 | 40.9 | 63.1 | 54.3 | 47.7 | — | 29.9 | 60.5 | |
| SpatialLadder-3BData Scale=26K, Training Phases=3 (S+R)2026.01 | 45.7 | — | — | — | — | — | — | — | — | |
| Gemini-1.5 ProData Scale=-, Training Phases=-2026.01 | 45.4 | 56.2 | 30.9 | 64.1 | 43.6 | 51.3 | 46.3 | 36 | 34.6 | |
| ViLaSR-7BData Scale=81K, Training Phases=2 (S+R)2026.01 | 45.4 | 63.5 | 34.4 | 60.6 | 30.9 | 48.9 | 45.2 | 30.4 | 49.2 | |
| SpaceR-7BData Scale=151K, Training Phases=2 (S+R)2026.01 | 43.5 | 61.9 | 28.6 | 60.9 | 35.2 | 38.2 | 46 | 31.4 | 45.6 | |
| LLaVA-Video-7BData Scale=-, Training Phases=-2026.01 | 35.6 | 48.5 | 14 | 47.8 | 24.2 | 43.5 | 42.4 | 34 | 30.6 | |
| GPT-4oData Scale=-, Training Phases=-2026.01 | 34 | 46.2 | 5.3 | 43.8 | 38.2 | 37 | 41.3 | 31.5 | 28.5 | |
| Qwen2.5-VL-7BData Scale=-, Training Phases=-2026.01 | 32.7 | 34.5 | 19.4 | 47.6 | 40.8 | 32.8 | 24.5 | 32.5 | 29.4 | |
| LLaVA-OneVision-7BData Scale=-, Training Phases=-2026.01 | 32.4 | 47.7 | 20.2 | 47.4 | 12.3 | 42.5 | 35.2 | 29.4 | 24.4 | |
| SAT-7BData Scale=175K, Training Phases=1 (S)2026.01 | — | — | — | — | — | 47.3 | 41.1 | 37.1 | 36.1 |