Benchmark RadarAI BENCHMARK PROFILE
Vector-Bench
Vector-Bench evaluates instruction-based SVG code editing with 40 repair tasks, deterministic specification rewards, and metrics including repair progress and unintended change rate, across 34 model endpoints.
- Released
- 2026-07-21
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Vector editing fidelity is under-evaluated; this benchmark provides precise specification metrics, but public availability is unclear.
Motivation
Instruction-based vector editing requires two capabilities: making a requested change and leaving everything else alone.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.