WildRoadBench - Guardrail Damage: leaderboard

The guardrail damage class of WildRoadBench (2026), an aerial road-damage grounding benchmark: models box every instance in wild drone images, 20 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 guardrail damage scene class (20 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: Around February 2027. 25 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro38.1#145
2Claude Sonnet 4.637.8#173
3Qwen3-VL-32B-Instruct-AWQ34.3
4Qwen 3.6 Plus27#238
5Kimi K2.525#266
6Gemini 3.1 Flash Lite23.8#371
7Qwen 3 VL 8B Instruct17.6#1029
8Qwen3-VL-32B-Thinking-AWQ17
9InternVL3-8B16.7#1782
10Qwen 3.5 VL 9B15.4
11Qwen 3.5 VL 4B11.4
12Gemini 2.5 Flash11.1#463
13GPT-5.4 (2026-03-05)10#92
14Qwen2.5-VL-32B-Instruct-AWQ5.7
15Qwen 3 VL 8B (Thinking)5.4#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-guardrail-damage · How It Works · Data refreshed daily, snapshot 2026-11-21.