Benchmark RadarCoinVE-Bench
CoinVE-Bench provides 361 multi-instruction test cases for compositional video editing, with four-dimensional evaluation metrics covering instruction following, editing accuracy, visual quality, and temporal consistency.
- Released
- 2026-08-18
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
CoinVE-Bench offers a dedicated evaluation set for compositional instruction-guided video editing, addressing gaps in existing single-operation benchmarks and enabling standardized model comparison.
Motivation
The quality and diversity of instruction-based video editing datasets are steadily improving, yet existing datasets mainly focus on single editing operations and fall short in supporting compositional instruction-guided video editing.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.