MamTiff-CAD: Multi-Scale Latent Diffusion with Mamba+ for Complex Parametric Sequence

2025-11-20

AI Summary

Addressing the challenge of generating long parametric CAD command sequences for complex models—constrained by geometric and topological dependencies—this paper proposes MamTiff-CAD, a novel autoencoder framework integrating Mamba+ (enhanced with a forget gate) and Transformer architectures. It constructs a multi-scale latent diffusion model to jointly capture long-range dependencies and structural constraints in the latent space. A non-autoregressive Transformer decoder is further designed for efficient sequence generation. To enable rigorous evaluation of long-sequence modeling, we introduce the first large-scale dataset of parametric CAD commands featuring extended sequences. Experiments demonstrate state-of-the-art performance in both reconstruction and generative tasks: the method stably produces high-quality command sequences of 60–256 steps, significantly enhancing controllability and fidelity in synthesizing complex CAD models.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Parametric Computer-Aided Design (CAD) is crucial in industrial applications, yet existing approaches often struggle to generate long sequence parametric commands due to complex CAD models' geometric and topological constraints. To address this challenge, we propose MamTiff-CAD, a novel CAD parametric command sequences generation framework that leverages a Transformer-based diffusion model for multi-scale latent representations. Specifically, we design a novel autoencoder that integrates Mamba+ and Transformer, to transfer parameterized CAD sequences into latent representations. The Mamba+ block incorporates a forget gate mechanism to effectively capture long-range dependencies. The non-autoregressive Transformer decoder reconstructs the latent representations. A diffusion model based on multi-scale Transformer is then trained on these latent embeddings to learn the distribution of long sequence commands. In addition, we also construct a dataset that consists of long parametric sequences, which is up to 256 commands for a single CAD model. Experiments demonstrate that MamTiff-CAD achieves state-of-the-art performance on both reconstruction and generation tasks, confirming its effectiveness for long sequence (60-256) CAD model generation.

Problem

• Generating long parametric CAD command sequences
• Capturing complex geometric and topological constraints
• Modeling distribution of long sequence CAD commands

Innovation

• Integrates Mamba+ and Transformer for latent representations
• Uses forget gate mechanism to capture long dependencies
• Employs multi-scale Transformer diffusion model for generation