AirGroundBench - Bearing Perception: leaderboard
AirGroundBench probes spatial intelligence for heterogeneous air-ground (UAV plus UGV) embodied collaboration, with nine multiple-choice VQA tasks in four capability dimensions plus a vision-language navigation task, evaluated zero-shot under synchronized dual views. This board reports Accuracy (%) on the bearing-perception VQA task; higher is better.
Metric: Accuracy (%) on the bearing-perception VQA task; higher is better. Source: arxiv.org. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 Plus | 48.28 | #232 |
| 2 | Qwen 3.5 397B A17B | 45.26 | #246 |
| 3 | Qwen 3.5 27B | 44.83 | #359 |
| 4 | Qwen 3 VL 235B A22B | 41.38 | #479 |
| 5 | Kimi K2.5 | 38.63 | #256 |
| 6 | Seed 2.0 Pro | 37.93 | #211 |
| 7 | ERNIE 5.0 | 37.93 | #456 |
| 8 | GLM-4.6V | 27.59 | #701 |
| 9 | Gemini 3.1 Pro (Preview) | 20.69 | #94 |
| 10 | GPT-5.2 | 20.69 | #176 |
| 11 | Claude Sonnet 4.6 | 17.24 | #161 |
Interactive version: theaggregate.ai/benchmark?slug=airgroundbench-bearing-perception · How It Works · Data refreshed daily, snapshot 2026-11-13.