SpatialQuery-1M - Proximity Decision: leaderboard
Part of SpatialQuery-1M (Nguyen et al., 2026): the model decides whether the closest same-category instance lies within a distance threshold of a reference object in an indoor RGB image, over 5,000 balanced queries at thresholds from 0.5 to 3 m. Accuracy in percent, credited only when the answer is correct on the image and not on a black image; higher is better.
Metric: Accuracy (%; binary decision whether the closest same-category instance lies within a distance threshold of the reference object, 5,000 fixed queries, 1,000 per threshold of 0.5, 1.0, 1.5, 2.0 and 3.0 m with balanced labels; black-image sanity filter: a query earns credit only when the model returns valid outputs for both the original RGB image and a same-size uniform black image, is correct on the original and incorrect on the black image; invalid outputs score zero; zero-shot, identical decoding settings). Source: arxiv.org. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 65.24 | #266 |
| 2 | GPT-4o | 60.26 | #741 |
| 3 | GPT-4o Mini | 54.46 | #1623 |
| 4 | Gemini 2.5 Flash | 52.52 | #453 |
| 5 | Qwen 2.5 VL 7B | 41.62 | #1787 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=spatialquery-1m-proximity-decision · How It Works · Data refreshed daily, snapshot 2026-11-13.