Benchmark RadarAI BENCHMARK PROFILE
GeoFidelity-Bench
GeoFidelity-Bench evaluates segment-level geographic fidelity in text-to-image street-view generation using 7,117 Mapillary images across 109 road segments in 25 cities.
- Released
- 2026-06-22
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
It addresses the gap between city-plausible and segment-accurate street-view generation, offering a protocol for comparing models on local geographic discrimination.
Motivation
Text-to-image models can generate visually plausible city streets, but whether their outputs correspond to a requested road segment rather than a generic city prior remains unclear.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.