Benchmark RadarAI BENCHMARK PROFILE
OmniCAD
Evaluates VLMs on assembly-aware 3D spatial reasoning across 25k mechanical assemblies with component-level, relational, and tool-augmented tasks.
- Released
- 2026-08-23
- Readiness
- Paper only
- Primary field
- Robotics & Autonomous Systems
Why it matters
Industrial assembly reasoning requires physically valid 3D understanding, and this benchmark exposes current VLM limitations in that domain.
Motivation
Recent vision-language models (VLMs) show strong capabilities in robotic perception and spatial reasoning, yet their ability to reason about complex mechanical assemblies remains underexplored.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.