UAVQA-Bench - Visual Grounding: leaderboard
Metric: Accuracy (%; simple, complex-semantic and highest-object grounding with a normalized bounding box, correct at IoU 0.5 or more; direct answering over UAVQA-Bench: 1,500 human-annotated questions on images from 13 public UAV datasets, 16 tasks in six capability dimensions, multiple choice plus box grounding counted correct at IoU 0.5). Source: arxiv.org. Saturation forecast: Around December 2026. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 82.55 |
| 2 | Gemini 3 Pro | 81.45 |
| 3 | Qwen 3 VL 8B Instruct | 80 |
| 4 | Qwen 3 VL 32B Instruct | 79.27 |
| 5 | Qwen 3 VL 32B (Thinking) | 79.27 |
| 6 | Qwen 3.5 9B | 78.55 |
| 7 | Qwen 3 VL 30B A3B Instruct | 75.64 |
| 8 | Qwen 3 VL 30B A3B (Thinking) | 50.91 |
| 9 | Qwen 3 VL 8B (Thinking) | 26.91 |
| 10 | InternVL3.5-8B | 3.63 |
Interactive version: theaggregate.ai/benchmark?slug=uavqa-bench-visual-grounding · How It Works · Data refreshed daily, snapshot 2026-09-29.