WildRoadBench - mAP: leaderboard
WildRoadBench (2026) asks vision-language models to box road damage in wild aerial images: water, cracks, potholes, debris and rare infrastructure defects. This board reports mAP averaged over IoU thresholds from 0.5 to 0.95, in percent, which rewards tight boxes; higher is better.
Metric: mAP (%) averaged over IoU thresholds 0.5 to 0.95, VLM Track, zero-shot box grounding on the full 1,061-image aerial set (1,394 ground-truth boxes); higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 25 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 14.7 | #145 |
| 2 | Qwen 3.6 Plus | 12.4 | #238 |
| 3 | Gemini 3.1 Flash Lite | 11.4 | #371 |
| 4 | Qwen 2.5 VL 7B Instruct | 9 | #1902 |
| 5 | Claude Sonnet 4.6 | 8.7 | #173 |
| 6 | Qwen2.5-VL-32B-Instruct-AWQ | 8.6 | |
| 7 | Qwen 3 VL 8B Instruct | 8.2 | #1029 |
| 8 | Qwen3-VL-32B-Instruct-AWQ | 7.9 | |
| 9 | Qwen3.5-VL-27B-FP8 | 7.8 | |
| 10 | Kimi K2.5 | 7 | #266 |
| 11 | GPT-5.4 (2026-03-05) | 6.9 | #92 |
| 12 | Qwen 3.5 VL 9B | 5.1 | |
| 13 | Gemini 2.5 Flash | 5 | #463 |
| 14 | Qwen3-VL-32B-Thinking-AWQ | 4.9 | |
| 15 | Qwen3.5-VL-35B-A3B-FP8 | 4.8 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=wildroadbench-map · How It Works · Data refreshed daily, snapshot 2026-11-21.