WildRoadBench - Guardrail Damage: leaderboard

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 May 2027. 25 models tracked.

Top models

#ModelScore
1Gemini 3 Pro38.1
2Claude Sonnet 4.637.8
3Qwen 3.6 Plus27
4Kimi K2.525
5Gemini 3.1 Flash Lite23.8
6Qwen 3 VL 8B Instruct17.6
7InternVL3-8B16.7
8Gemini 2.5 Flash11.1
9GPT-5.4 (2026-03-05)10
10Qwen 3 VL 8B (Thinking)5.4
11Claude Opus 4.64
12InternVL2-8B0
13InternVL3-14B0
14GPT-4o (2024-05-13)0
15Qwen 2.5 VL 7B Instruct0

Interactive version: theaggregate.ai/benchmark?slug=wildroadbench-guardrail-damage · How It Works · Data refreshed daily, snapshot 2026-10-07.