Benchmark RadarAI BENCHMARK PROFILE
EASEL
Evaluates dexterous visual tool use through reference-guided painting, semantic annotation, handwriting, and path planning tasks.
- Released
- 2026-08-26
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Introduces closed-loop, parameterized visual action as an underexplored agent capability beyond static QA and navigation.
Motivation
Evaluation is shifting from static QA toward agentic settings where models act through external tools.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.