ArtiCAD: Articulated CAD Assembly Design via Multi-Agent Code Generation

arXiv 2026 58.8 method

TLDR

ArtiCAD is a training-free multi-agent system that generates editable, articulated CAD assemblies from text or images using specialized agents and a connector for joint parameters.

Reasoning

The paper introduces a novel approach to articulated CAD assembly generation, addressing an unexplored area with a multi-agent framework that avoids training. Strengths include the use of validation steps, rollback mechanisms, and an experience store, while weaknesses are the lack of detailed quantitative results in the abstract and potential reliance on LLM/VLM limitations.

Read-first score

Read-first score 58.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 73.

Recency 8%

100

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

Methodology quality 25%

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

Reproducibility 25%

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

Topical relevance 42%

45.6

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 8.

Keyword Scores

computer-aided design

CAD generation

text-to-CAD

image-to-CAD

generative CAD

parametric CAD

AI for CAD

CAD program

neural CAD

CAD reconstruction

B-Rep

boundary representation

constructive solid geometry

CSG

sketch extrusion

point cloud to CAD

Deep Analysis

Innovations

First training-free multi-agent system for generating editable, articulated CAD assemblies from text or images.
Division of task among four specialized agents: Design, Generation, Assembly, and Review.
Predicting assembly relationships during initial design stage via a Connector that defines attachment points and joint parameters before geometry generation, bypassing limited spatial reasoning of LLMs/VLMs.
Validation steps in generation and assembly stages with cross-stage rollback mechanism to isolate and correct design- and code-level errors.
Self-evolving experience store that accumulates design knowledge to continuously improve performance.

Methodology

ArtiCAD is a multi-agent system with four agents (Design, Generation, Assembly, Review) that generates articulated CAD assemblies from text or images. It uses a Connector to pre-define assembly relationships before geometry generation, incorporates validation and cross-stage rollback for error correction, and maintains a self-evolving experience store. The system is evaluated on three datasets: ArtiCAD-Bench, CADPrompt, and ACD.

Key Results

Extensive evaluations on three datasets (ArtiCAD-Bench, CADPrompt, ACD) validate the effectiveness of ArtiCAD, and its applicability is demonstrated in requirement-driven conceptual design, physical prototyping, and URDF export for embodied AI training assets.

Resources

Paper