AeroGround - Occluded Target Judgment: leaderboard
Metric: Accuracy (%; Spatial Reasoning: decide whether a target is visible, occluded or outside the field of view; aerial-ground collaborative questions from paired simulated UAV and ground views, exact-match scoring of the parsed option or mapping, invalid output scored wrong; one deterministic run). Source: arxiv.org. Saturation forecast: Around 2034. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Seed 2.0 Pro | 45 |
| 2 | Qwen 3 VL 8B Instruct | 42.8 |
| 3 | GPT-5.4 | 42.2 |
| 4 | Qwen 3 VL 32B Instruct | 38.8 |
| 5 | Kimi K2.5 | 38.1 |
| 6 | Qwen 3 VL 32B (Thinking) | 37.5 |
| 7 | Gemini 3.1 Pro (Preview) | 36.6 |
| 8 | Qwen 3 VL 8B (Thinking) | 28.4 |
Interactive version: theaggregate.ai/benchmark?slug=aeroground-occluded-target-judgment · How It Works · Data refreshed daily, snapshot 2026-09-26.