NeSy-Route - Region Matching: leaderboard
Metric: Region matching rate (%): accuracy of the region vector that assigns each land-cover class present in the image to its annotated region, on NeSy-Route Task 2 (text-image constraint alignment, 12,975 remote-sensing images in Easy, Medium and Hard tiers by number of land-cover classes): the model grounds the query's constraints in identified image regions; mean of the three tier scores; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 71.01 | #77 |
| 2 | Qwen 3.5 27B | 60.56 | #183 |
| 3 | Qwen 3 VL 30B A3B Instruct | 57.31 | #365 |
| 4 | GPT-5.1 | 56.26 | #131 |
| 5 | Qwen 3 VL 32B Instruct | 48.02 | #276 |
| 6 | Qwen 3 VL 235B A22B Instruct | 43.66 | #264 |
| 7 | Qwen 3 VL 8B Instruct | 22.01 | #401 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=nesy-route-region-matching · How It Works · Data refreshed daily, snapshot 2026-10-11.