WildRoadBench - Pavement Debris: leaderboard

The pavement debris class of WildRoadBench (2026), an aerial road-damage grounding benchmark: models box every instance in wild drone images, 85 ground-truth boxes in all. The score is F1 at IoU 0.5 in percent; higher is better.

Metric: F1 (%) of predicted boxes at IoU 0.5 for the pavement debris scene class (85 ground-truth boxes), 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: Not forecast. 25 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro57.3#145
2Qwen2.5-VL-32B-Instruct-AWQ45.5
3Claude Sonnet 4.641.4#173
4Qwen 3.6 Plus41.2#238
5Qwen 2.5 VL 7B Instruct40.6#1902
6Qwen3-VL-32B-Instruct-AWQ33.5
7Gemini 3.1 Flash Lite32.8#371
8Qwen 3.5 VL 4B28.7
9Qwen 3 VL 8B Instruct27.4#1029
10Qwen 3.5 VL 9B26.7
11Qwen3.5-VL-35B-A3B-FP826.5
12Qwen3.5-VL-27B-FP824.2
13Kimi-VL-A3B-Instruct21.1#2253
14Qwen3-VL-32B-Thinking-AWQ20.5
15Qwen 3 VL 8B (Thinking)18.6#910

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=wildroadbench-pavement-debris · How It Works · Data refreshed daily, snapshot 2026-11-21.