Benchmark RadarAI BENCHMARK PROFILE
AV-SyncBench
AV-SyncBench evaluates audio-visual synchronization by separating temporal and semantic consistency. It contains 3,269 videos and 38,390 samples across 10 scenarios and 5 tasks, with data filtered and verified for on-screen sound sources.
- Released
- 2026-07-01
- Readiness
- Inspectable
- Primary field
- General AI
Why it matters
Existing AV feature extraction evaluations are coupled, preventing independent assessment of temporal and semantic alignment. AV-SyncBench provides a decoupled benchmark to quantify feature quality for alignment and downstream tasks.
Motivation
Audio-visual feature extraction is a fundamental component of multimodal understanding and generation tasks.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.