VST-7B-SFT: benchmark results
Provider: Other.
Aggregate Rating 1599 ± 30, rank #1020 of 10916 rated models, from 45 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Metric-Bench | 39.61 | Mean relative accuracy (%; all questions; 1,340 template que | 100 |
| Metric-Bench - Size | 42.23 | Mean relative accuracy (%; length, width, height, elevation | 100 |
| SpatialUAV - Occlusion Removal | 46 | Occlusion Removal Accuracy (%; exact region IDs of targets h | 94.1 |
| SpatialUAV - Global Motion | 54.4 | Motion Description Similarity (%; GPT-5.4-mini judge, 0-1 pe | 88.2 |
| CourtSI-Bench - Player-Player | 65.9 | Exact-match accuracy (%) on the 393 player-to-player relatio | 86.4 |
| MindEdit-Bench - Cross-View Perspective Transformation | 12 | Accuracy (%) | 85.7 |
| MindEdit-Bench - Single-View Perspective Transformation | 15 | Accuracy (%) | 85.7 |
| CourtSI-Bench - Camera-Player | 78.6 | Exact-match accuracy (%) on the 248 camera-to-player relatio | 82.6 |
| Metric-Bench - Depth | 53.84 | Mean relative accuracy (%; object-depth questions; 1,340 tem | 78.6 |
| CourtSI-Bench - Player-Zone | 48.8 | Exact-match accuracy (%) on the 82 player-to-court-zone rela | 75 |
| CourtSI-Bench | 39.6 | Overall score (%): the question-weighted mean of the 13 task | 68.2 |
| CourtSI-Bench - Player-Line | 61 | Exact-match accuracy (%) on the 495 player-to-court-line rel | 68.2 |
Interactive version: theaggregate.ai/model?slug=vst-7b-sft · How It Works · Data refreshed daily, snapshot 2026-11-08.