# Learn the Solid, Not the File: Canonical Inputs for Neural Networks on CAD Boundary Representations

arXiv: https://arxiv.org/abs/2609.11573 · HTML: https://mlsanity.com/paper/2609.11573/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 47 references that are in the map.
- First posted on arXiv 10 Sep 2026.

## Builds on (most cited first, 25 of 25)

- [Dynamic Graph CNN for Learning on Point Clouds](https://mlsanity.com/paper/1801.07829/) (2018)
- [DeepCAD: A Deep Generative Network for Computer-Aided Design Models](https://mlsanity.com/paper/2105.09492/) (2021)
- [BrepGen: A B-rep Generative Diffusion Model with Structured Latent Geometry](https://mlsanity.com/paper/2401.15563/) (2024)
- [BRepNet: A topological message passing system for solid models](https://mlsanity.com/paper/2104.00706/) (2021)
- [SolidGen: An Autoregressive Model for Direct B-rep Synthesis](https://mlsanity.com/paper/2203.13944/) (2022)
- [JoinABLe: Learning Bottom-up Assembly of Parametric CAD Joints](https://mlsanity.com/paper/2111.12772/) (2021)
- [CAD-Recode: Reverse Engineering CAD Code from Point Clouds](https://mlsanity.com/paper/2412.14042/) (2024)
- [HoLa: B-Rep Generation using a Holistic Latent Representation](https://mlsanity.com/paper/2504.14257/) (2025)
- [DTGBrepGen: A Novel B-rep Generative Model through Decoupling Topology and Geometry](https://mlsanity.com/paper/2503.13110/) (2025)
- [Churn Reduction via Distillation](https://mlsanity.com/paper/2106.02654/) (2021)
- [AutoBrep: Autoregressive B-Rep Generation with Unified Topology and Geometry](https://mlsanity.com/paper/2512.03018/) (2025)
- [Geometric Deep Learning for Computer-Aided Design: A Survey](https://mlsanity.com/paper/2402.17695/) (2024)
- [BrepGPT: Autoregressive B-rep Generation with Voronoi Half-Patch](https://mlsanity.com/paper/2511.22171/) (2025)
- [Pointer-CAD: Unifying B-Rep and Command Sequences via Pointer-based Edges & Faces Selection](https://mlsanity.com/paper/2603.04337/) (2026)
- [BRepFormer: Transformer-Based B-rep Geometric Feature Recognition](https://mlsanity.com/paper/2504.07378/) (2025)
- [BRepMAE: Self-Supervised Masked BRep Autoencoders for Machining Feature Recognition](https://mlsanity.com/paper/2602.22701/) (2026)
- [B-rep Matching for Collaborating Across CAD Systems](https://mlsanity.com/paper/2306.03169/) (2023)
- [BrepCoder: A Unified Multimodal Large Language Model for Multi-task B-rep Reasoning](https://mlsanity.com/paper/2602.22284/) (2026)
- [FoV-Net: Rotation-Invariant CAD B-rep Learning via Field-of-View Ray Casting](https://mlsanity.com/paper/2602.24084/) (2026)
- [Masked BRep Autoencoder via Hierarchical Graph Transformer](https://mlsanity.com/paper/2603.14927/) (2026)
- [DualBrep: A Dual-Field Continuous Representation for B-rep Modelling](https://mlsanity.com/paper/2606.31579/) (2026)
- [Autoregressive B-Rep Shape Generation with Parametric Surfaces](https://mlsanity.com/paper/2607.17093/) (2026)
- [CAD-feature enhanced machine learning for manufacturing effort estimation on sheet metal bending parts](https://mlsanity.com/paper/2605.12266/) (2026)
- [BRepCLIP: Contrastive Multimodal Pretraining on BRep Primitives for CAD Understanding](https://mlsanity.com/paper/2606.05515/) (2026)
- [Masked Topology Modeling for Self-Supervised Learning on Parametric CAD](https://mlsanity.com/paper/2607.20642/) (2026)

## Abstract (arXiv)

Boundary representation (B-rep) is the standard format used by modern CAD systems for parametric 3D models. It turns out, the exact same solid can be represented by different B-reps: for example, two engineers using different operations, a geometry kernel rebuilding the file, and an export setting repartitioning faces will lead to different B-reps even though the underlying solid remains the same. We show that existing B-rep encoders are not robust to variation in the B-rep with the same solid on perturbations applied to standard benchmarks, naturally occurring variations inherent to CAD software, and differences in how designers model the same part via a human dataset we created in FreeCAD. The performance of popular B-rep encoders often collapses catastrophically. We propose the canonical region graph, an input representation whose nodes, features and coordinate frame are derived from the solid itself and show theoretical invariance guarantees on repartitioning and rigid motions. It matches the strongest baseline on standard benchmarks, and is stable under every perturbation we test.