Open Spatial Reasoning: Multiple-Choice 3D Reasoning Questions from Driving Images

Open Spatial Reasoning is a multiple-choice dataset for evaluating 3D spatial reasoning from single driving images. The dataset was created by ReasonCore and was last updated on 2026-05-29. Each image contains numbered bounding boxes referencing objects, and questions probe a model's ability to reconstruct the real 3D scene.

Use Cases

• Benchmarking 3D spatial reasoning in AI models based on single driving images.
• Training models to avoid flat-image shortcuts like 'lower in the frame = closer'.
• Evaluating scene reconstruction capabilities using images with numbered bounding boxes.

Strengths

• Designed to evaluate 3D spatial reasoning, a specific and challenging AI capability.
• Questions probe for real 3D scene reconstruction rather than reliance on 2D image heuristics.
• Images include numbered bounding boxes to reference specific objects in the scene.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown.

Provenance

Source

ReasonCore

Freshness

Last updated 2026-05-29 16:19:16; freshness should be verified.