WildRoadBench - Drainage Damage: leaderboard
Metric: F1 (%) of predicted boxes at IoU 0.5 for the drainage damage scene class (22 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 March 2027. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.6 Plus | 58.3 |
| 2 | Gemini 3 Pro | 57.8 |
| 3 | Gemini 3.1 Flash Lite | 50 |
| 4 | Qwen 3 VL 8B Instruct | 45.5 |
| 5 | Claude Sonnet 4.6 | 42.3 |
| 6 | Kimi K2.5 | 35.1 |
| 7 | Qwen 2.5 VL 7B Instruct | 23.8 |
| 8 | Gemini 2.5 Flash | 22.7 |
| 9 | GPT-5.4 (2026-03-05) | 21.7 |
| 10 | InternVL3-8B | 18.2 |
| 11 | GPT-4o (2024-05-13) | 18.2 |
| 12 | InternVL3-14B | 17.8 |
| 13 | Claude Opus 4.6 | 13.5 |
| 14 | Qwen 3 VL 8B (Thinking) | 12.2 |
| 15 | InternVL2-8B | 0 |
Interactive version: theaggregate.ai/benchmark?slug=wildroadbench-drainage-damage · How It Works · Data refreshed daily, snapshot 2026-10-07.