Benchmark RadarV2V-Bench
V2V-Bench evaluates video-to-video generation models across 11 dimensions in five categories: temporal alignment, structural fidelity, transformation quality, video quality, and semantic alignment. It pairs source videos with editing tasks and scores models on these dimensions.
- Released
- 2026-06-04
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing T2V and I2V metrics do not capture the joint requirements of instruction following and frame-level correspondence in V2V generation. A dedicated benchmark with human-correlated scoring could support model selection for V2V applications.
Motivation
Video-to-video (V2V) generation is difficult to evaluate because outputs must both follow editing instructions and preserve frame-level correspondence with the source video, which existing T2V and I2V metrics do not capture.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.