AI BENCHMARK PROFILE
VIBE V2
VIBE-V2 is an internal benchmark covering pure front-end and full-stack Web, Android, and iOS projects with build-from-scratch tasks. It uses an Agent-as-a-Verifier paradigm to automatically verify program interaction logic and visual output, scoring models through a unified pipeline that includes a requirement set, containerized deployment, and a dynamic interaction environment.
- Released
- Unknown
- Readiness
- Paper only
- Primary field
- General AI
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.