Benchmark RadarAI BENCHMARK PROFILE
CraftBench
CraftBench evaluates scientific figure generation across three figure types and four input conditions, with human-drawn targets and a referenced VLM judge for scoring.
- Released
- 2026-05-28
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Automated figure generation lacks comprehensive benchmarks covering diverse types and conditions. CraftBench provides a standardized evaluation to measure progress in editable scientific figure generation.
Motivation
Scientific figures are among the most effective means of communicating complex research ideas, yet producing publication-quality illustrations remains one of the most labor-intensive parts of paper preparation.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.