Benchmark RadarSCOPE-Bench
SCOPE-Bench evaluates short-video recommendation systems by quantifying content depth using the Content Depth Score (CDS), a seven-level scale based on cognitive psychology. It provides CDS annotations for 150K videos from an open-source dataset, enabling systematic assessment of recommenders from a cognitive-content perspective.
- Released
- 2026-08-14
- Readiness
- Inspectable
- Primary field
- General AI
Why it matters
Existing short-video recommenders optimize for engagement, often favoring shallow content. SCOPE-Bench addresses the lack of benchmarks for content depth in recommendation, allowing evaluation of algorithms on their ability to recommend cognitively deep content, which has implications for user well-being and long-term engagement.
Motivation
Driven by the attention economy, short-video Recommender Systems (RSs) are primarily optimized to maximize user engagement by promoting videos that capture attention within seconds.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.