UVLM Street-View Benchmark (Reasoning): leaderboard
Metric: Mean proximity score (%) in advanced-reasoning mode (chain of thought before the answer, at most 1,024 new tokens): the unweighted mean of five proximity scores on 120 street-level photographs of French urban frontages scored against one expert annotator (motor-vehicle count, sidewalk presence, pedestrian-entrance count, street-frontage length in metres, vegetation type 1-6); proximity is 100 minus the absolute error as a percentage of the observed range of human values; answers are the majority of 3 greedy runs (consensus validation) through the UVLM loader; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 32B Instruct | 88 | #443 |
| 2 | Qwen 2.5 VL 7B Instruct | 77.4 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=uvlm-street-view-benchmark-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-11.