Benchmark RadarP3D-Bench
P3D-Bench evaluates multimodal large language models on parametric 3D generation from text, image, and assembly specifications, scoring executability, geometric fidelity, topology, text-grounded constraints, multiview semantic alignment, and part-level structure.
- Released
- 2026-06-09
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Existing benchmarks rarely evaluate 3D modeling through code, which requires geometric precision and assembly consistency, not just runnable code. P3D-Bench provides a unified protocol to assess structural understanding and precise geometry, which is critical for models generating parametric 3D programs.
Motivation
Multimodal large language models can write code to produce complex programs as well as use programs to do 3D modeling, which opens up a new avenue for 3D generation powered by their priors, world knowledge and reasoning.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.