SpatialUAV - Object Matching: leaderboard
Object matching across two UAV views from SpatialUAV (Zhang et al., 2026): the model locates a target from one view in the other, answering with a bounding box that counts when a combined overlap, center and size score reaches 0.5, or with a region ID that must match exactly. Accuracy in percent over 212 questions; higher is better.
Metric: Object Matching Accuracy (%; box accepted at a composite IoU, center and size score of at least 0.5, or exact region ID). Source: arxiv.org. Saturation forecast: Around June 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 28.8 | #121 |
| 2 | Claude Opus 4.7 | 22.2 | #75 |
| 3 | Qwen 3.6 35B A3B | 10.4 | #461 |
| 4 | Qwen 3.6 27B | 9 | #339 |
| 5 | Qwen 3.5 9B | 7.1 | #803 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=spatialuav-object-matching · How It Works · Data refreshed daily, snapshot 2026-11-12.