WildRoadBench - Drainage Water: leaderboard
Metric: F1 (%) of predicted boxes at IoU 0.5 for the drainage water scene class (24 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.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Flash | 16.4 |
| 2 | Qwen 2.5 VL 7B Instruct | 11.8 |
| 3 | Qwen 3 VL 8B Instruct | 10.2 |
| 4 | Gemini 3.1 Flash Lite | 9.2 |
| 5 | InternVL3-14B | 7.8 |
| 6 | Qwen 3 VL 8B (Thinking) | 7.3 |
| 7 | GPT-5.4 (2026-03-05) | 6.6 |
| 8 | Qwen 3.6 Plus | 5 |
| 9 | InternVL3-8B | 4.3 |
| 10 | Gemini 3 Pro | 3.3 |
| 11 | Claude Sonnet 4.6 | 2.8 |
| 12 | Claude Opus 4.6 | 0 |
| 13 | Kimi K2.5 | 0 |
| 14 | InternVL2-8B | 0 |
| 15 | GPT-4o (2024-05-13) | 0 |
Interactive version: theaggregate.ai/benchmark?slug=wildroadbench-drainage-water · How It Works · Data refreshed daily, snapshot 2026-10-07.