Benchmark RadarAI BENCHMARK PROFILE
GeoT2V-Bench
GeoT2V-Bench evaluates 3D consistency in camera-prompted text-to-video models via 3D reconstruction, using metrics like static rendering error and flow agreement.
- Released
- 2026-06-23
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Assesses whether generated videos can support explicit rigid 3D reconstruction, providing a diagnostic tool for a critical limitation of T2V models.
Motivation
Camera-prompted text-to-video (T2V) models are increasingly used to synthesize virtual camera captures, such as orbiting objects or moving through static scenes.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.