Foundation Models for Automatic CAD Generation

用于自动 CAD 生成的基础模型

Recent advances in Large Language Models (LLMs) and Vision-Language Models (VLMs) enable the automatic generation of parametric 3D designs from natural-language specifications. This chapter presents an empirical study of foundation models for automatic Computer-Aided Design (CAD) generation of mechanical parts, using a unified evaluation pipeline and a curated benchmark of 97 engineering design problems.

大型语言模型（LLM）和视觉语言模型（VLM）的最新进展，使得根据自然语言规范自动生成参数化 3D 设计成为可能。本章对用于机械零件自动计算机辅助设计（CAD）生成的基础模型进行了实证研究，采用了统一的评估流程，并使用了一个包含 97 个工程设计问题的精选基准测试集。

We introduce LLMForge, a multi-model text-to-CAD framework integrating JSON-schema validation, analytic feature scoring, mesh synthesis, and multi-round iterative refinement, studied under two critique regimes. IterTracer uses a Phong-shaded ray-trace renderer with analytic visual metrics (silhouette IoU, hole visibility, edge clearance, aspect-ratio conformance) for lightweight geometry-aware feedback across rounds. IterVision replaces the analytic scorer with a VLM semantic critic (Qwen2.5-VL-72B) that evaluates rendered views via chain-of-thought visual reasoning, assessing spatial coherence and design intent.

我们引入了 LLMForge，这是一个集成了 JSON 模式验证、解析特征评分、网格合成和多轮迭代优化的多模型文本转 CAD 框架，并在两种评估机制下进行了研究。IterTracer 使用带有 Phong 着色光线追踪渲染器和解析视觉指标（轮廓 IoU、孔可见性、边缘间隙、纵横比一致性）的工具，在各轮次中提供轻量级的几何感知反馈。IterVision 则将解析评分器替换为 VLM 语义评估器（Qwen2.5-VL-72B），通过思维链视觉推理评估渲染视图，从而判断空间连贯性和设计意图。

On a benchmark spanning four canonical geometry families (plates with holes and bolt circles, multi-feature boxes, flanged cylinders, and L-brackets), we evaluate seven foundation models: DeepSeek-V3.2, Qwen3-235B-A22B, Llama-3.3-70B, Gemma-3-27B, GLM-4.5, MiniMax-M2.1, and INTELLECT. Under IterTracer, the four highest-ranked models form a tight cluster (overall mean in [0.885, 0.890]) with 98.97% mesh success, showing that compact instruction-tuned models can match substantially larger systems.

在涵盖四种典型几何族（带孔和螺栓圆的板、多特征盒、法兰圆柱体和 L 型支架）的基准测试中，我们评估了七个基础模型：DeepSeek-V3.2、Qwen3-235B-A22B、Llama-3.3-70B、Gemma-3-27B、GLM-4.5、MiniMax-M2.1 和 INTELLECT。在 IterTracer 机制下，排名最高的四个模型形成了一个紧密的集群（总体平均分在 [0.885, 0.890] 之间），网格生成成功率达到 98.97%，这表明紧凑的指令微调模型完全可以媲美规模大得多的系统。

VLM-based critique in IterVision yields 100% watertight mesh generation on the leading model while surfacing systematic difficulty on rotationally symmetric geometries such as cylinders, where visual and semantic scoring diverge most. We discuss benchmark design, failure modes, CAD-oriented prompting, and implications for industrial workflows and scalable automated mechanical design.

IterVision 中基于 VLM 的评估机制使领先模型实现了 100% 的封闭网格生成，同时也揭示了在旋转对称几何体（如圆柱体）上存在的系统性难题，在这些几何体上，视觉评分与语义评分的分歧最为显著。我们讨论了基准测试设计、故障模式、面向 CAD 的提示工程，以及对工业工作流程和可扩展自动化机械设计的启示。