Benchmark RadarAI BENCHMARK PROFILE
VES-Bench
Tests long-horizon video understanding with 600 Temporal Ordering and Event Counting questions, auditing whether decoded frames cover all evidence intervals.
- Released
- 2026-08-23
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Shifts evaluation from final answers to evidence coverage, revealing whether correct answers rest on complete observation of long videos.
Motivation
A long-video answer is evidence-supported only when the frames decoded from the video cover every event the answer depends on.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.