Benchmark RadarAI BENCHMARK PROFILE
SpatialGen-Bench
ProVisE is a framework for evaluating image-generation models on spatial benchmarks by converting visual answers into structured predictions. SpatialGen-Bench is a diagnostic dataset of 470 samples across 14 spatial subtasks.
- Released
- 2026-07-23
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Existing spatial benchmarks rely on text or coordinates, limiting image-generation models. ProVisE adapts visual answers to original metrics, enabling comparison. However, unclear if it is a standalone benchmark or a framework.
Motivation
Spatial intelligence is essential for agents to move from static semantic understanding toward interacting with the physical world.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.