SpatialUAV - Occlusion Removal: leaderboard
Occlusion reasoning across two UAV views from SpatialUAV (Zhang et al., 2026): the model identifies targets that are visible in one view but hidden in the other, answering with region IDs that must match the annotation exactly. Accuracy in percent over 63 questions; higher is better.
Metric: Occlusion Removal Accuracy (%; exact region IDs of targets hidden in one of two views). Source: arxiv.org. Saturation forecast: Around 2031. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 49.2 | #121 |
| 2 | Qwen 3.5 9B | 33.3 | #803 |
| 3 | Qwen 3.6 27B | 33.3 | #339 |
| 4 | Qwen 3.6 35B A3B | 30.2 | #461 |
| 5 | Claude Opus 4.7 | 22.2 | #75 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=spatialuav-occlusion-removal · How It Works · Data refreshed daily, snapshot 2026-11-12.