CVSBench - CVUSA Ground-to-Satellite: leaderboard
CVSBench (2026) tests cross-view spatial reasoning with paired street-view and satellite images: models infer attributes of the unseen view and localize the street-view camera on the satellite image, on four-choice and binary questions. This board is accuracy in percent on CVUSA-subset questions that infer satellite-view attributes from a street-view image; higher is better.
Metric: Accuracy (%) on CVUSA-subset ground-to-satellite questions: from a street-view image, infer spatial attributes of the unseen satellite view; four-choice and binary-choice questions over paired street-view and satellite images (Gemini-2.5-flash-generated, human-verified), test split; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 4B (IT) | 71 | #3831 |
| 2 | Qwen 3 VL 8B Instruct | 69.9 | #981 |
| 3 | GPT-5 Chat | 69.7 | #444 |
| 4 | Qwen 3 VL 4B Instruct | 67.9 | #1279 |
| 5 | DeepSeek V3.2 | 63.39 | #406 |
| 6 | GPT-4o | 51.8 | #741 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=cvsbench-cvusa-ground-to-satellite · How It Works · Data refreshed daily, snapshot 2026-11-13.