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

Towards High-Fidelity CAD Generation via LLM-Driven Program Generation and Text-Based B-Rep Primitive Grounding

arXiv 2026 74.4 method

TLDR

FutureCAD uses LLMs and B-Rep grounding to generate high-fidelity CAD scripts from text, achieving state-of-the-art performance with a real-world dataset.

Reasoning

The paper introduces a novel integration of parametric modeling and B-Rep via LLM-driven program generation and a grounding transformer, supported by a new real-world dataset and reinforcement learning. Strengths include the unified framework and empirical results, but the abstract lacks detailed evaluation metrics and comparisons to prior methods.

Read-first score

Read-first score 74.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 105.

Recency 8%
100

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

Reproducibility 25%
85

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

Methodology quality 25%
70

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

Topical relevance 42%
65.6

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 anchorReproducibility anchor

Rank sensitivity

Stability: stable; rank range: 1.

Keyword Scores

B-Rep
10
boundary representation
10
CAD generation
10
text-to-CAD
10
AI for CAD
9
computer-aided design
9
generative CAD
9
CAD program
9
parametric CAD
8
neural CAD
7
sketch extrusion
4
constructive solid geometry
3
CSG
3
CAD reconstruction
2
image-to-CAD
1
point cloud to CAD
1

Deep Analysis

Innovations

  • LLM-driven generation of executable CadQuery scripts for CAD modeling
  • Text-based B-Rep primitive grounding via a transformer (BRepGround) that interprets natural language geometric selections
  • Unified framework bridging parametric CAD modeling and direct B-Rep synthesis
  • Training pipeline combining supervised fine-tuning and reinforcement learning for improved generalization

Methodology

FutureCAD uses a large language model to generate CadQuery scripts from text prompts, and a B-Rep grounding transformer to map natural language queries to geometric primitives. The LLM is first supervised fine-tuned on a new real-world CAD dataset, then further optimized with reinforcement learning.

Key Results

FutureCAD achieves state-of-the-art performance on CAD generation tasks.

Tags

CV