Awesome AI4CAD Hub Papers · Datasets · Projects
← Back to papers

CADEvolve: Creating Realistic CAD via Program Evolution

arXiv 2026 65.3 method

TLDR

CADEvolve uses VLM-guided evolution to generate complex CAD programs, creating a large dataset and achieving SOTA on Image2CAD.

Reasoning

The paper addresses the data bottleneck in CAD automation by evolving simple primitives into complex programs, validated through real benchmarks. Strengths include a novel evolution pipeline and large dataset; weaknesses include limited explicit discussion of generalization beyond image-to-CAD tasks.

Read-first score

Read-first score 65.3, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 96.

Recency 8%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 25%
90

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,dataset,metric,result,validation

Topical relevance 42%
60

Uses existing LLM keyword relevance scores normalized to 0-100. AI for CAD,computer-aided design,neural CAD,generative CAD,parametric CAD,B-Rep,boundary representation,constructive solid geometry,CSG,sketch extrusion,CAD generation,CAD reconstruction,text-to-CAD,image-to-CAD,point cloud to CAD,CAD program

Reproducibility 25%
38

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=dataset

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 9.

Keyword Scores

image-to-CAD
10
AI for CAD
9
generative CAD
9
parametric CAD
9
CAD generation
9
CAD program
9
computer-aided design
8
neural CAD
8
CAD reconstruction
8
sketch extrusion
5
constructive solid geometry
3
CSG
3
B-Rep
2
boundary representation
2
text-to-CAD
1
point cloud to CAD
1

Deep Analysis

Innovations

  • Evolution-based pipeline (CADEvolve) that incrementally grows CAD programs from simple primitives via VLM-guided edits and validations to achieve industrial-grade complexity.
  • Dataset of 8k complex parts as executable CadQuery parametric generators, post-processed and augmented to 1.3m scripts paired with rendered geometry, covering the full CadQuery operation set.
  • Fine-tuning a VLM on this dataset yields state-of-the-art Image2CAD performance across multiple benchmarks.

Methodology

CADEvolve starts from simple primitives and uses a VLM to guide iterative edits and validations, evolving CAD programs into complex, realistic parts expressed in CadQuery. The resulting 8k programs are post-processed and augmented to 1.3m script–geometry pairs, then used to fine-tune a VLM for the Image2CAD task.

Key Results

A VLM fine-tuned on the CADEvolve dataset achieves state-of-the-art results on the Image2CAD task across DeepCAD, Fusion 360, and MCB benchmarks.

Tags

GR