Benchmark RadarAI BENCHMARK PROFILE
CV-Arena
CV-Arena is a benchmark of 12K high-resolution image instruction pairs across 16 task types, with a human-AI collaborative preference protocol (Active Elo) for evaluating instructional computer vision problem solving.
- Released
- 2026-05-30
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses the gap in evaluating diverse real-image editing tasks with professional constraints, providing a scalable and traceable evaluation resource for model comparison.
Motivation
Instruction-guided image editing is becoming a general interface for visual work, yet existing benchmarks still focus largely on narrow appearance edits and do not fully capture the diversity of real-image tasks in professional workflows.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.