Text2CAD-Bench
600 human-curated NL-to-parametric-CAD examples in four complexity levels with layperson and expert dual-style prompts.
What it does and how it connects
Text2CAD-Bench is a benchmark for LLM-based text-to-parametric-CAD generation. It comprises 600 human-curated examples spanning four levels: L1 and L2 cover fundamental geometry with standard features, L3 introduces complex topology and freeform surfaces, and L4 extends to real-world domains beyond mechanical parts.
Each example pairs dual-style prompts: geometric descriptions mimicking non-expert users, and procedural sequences aligned with expert-level conventions. The paper positions it as the first benchmark systematically evaluating text-to-CAD across geometric complexity and application diversity, where earlier benchmarks focused on basic primitives and simple sketch-extrude sequences.
Evaluating mainstream general LLMs and domain-specific models, the paper finds current models perform reasonably on basic geometry but degrade substantially on complex topology and advanced features.
Alternatives
Compare side by side- Text2CAD
A framework and dataset of around 170K models that generates sketch-and-extrude sequences from text, rather than a benchmark that scores text-to-CAD generation on 600 curated examples.
- CADBench (Doris et al.)
A unified 18,000-sample benchmark turning mesh, image or render inputs into STEP or mesh, rather than 600 human-curated text-to-CAD examples across four complexity levels.
- BenchCAD
Scores execution-verified CadQuery programs by deterministic voxel IoU across 106 industrial part families, rather than 600 curated text-to-CAD examples across four complexity levels.
- MUSE
Grades text-to-CAD assemblies in a three-stage funnel ending in a VLM judge, rather than 600 curated examples across four complexity levels with dual-style prompts.
- DeepCAD (dataset)
A dataset of 178K sketch-and-extrude sequences parsed from Onshape that text-to-CAD work builds on, rather than a benchmark that scores text-to-CAD generation.
Sources
Not verified
- License: no license is stated on the arXiv page read.
- A public repo or dataset link: the abstract states the benchmark is released, but no repository or dataset URL appears on the abstract page read.
- Pricing: none stated (research benchmark).
- Real user voices: no real linked post read for this page.
- Related Error Index entries: no matching entry verified.
Last verified November 6, 2026. Listings are informational. The evidence score measures public, verifiable evidence and is not an endorsement or a compatibility claim.