acceptodds
Under review as a conference paper at ICLR 2027

Masked Topology Modeling for Self-Supervised Learning on Parametric CAD

Abstract

Computer aided design (CAD) is ubiquitous: virtually any modern object was designed using editable CAD tools. However, with the shortage of available CAD datasets in its native editable and parametric format, boundary representation (B-Rep), it is ever more important to develop data-efficient methods for this domain. We present a new self-supervised pretraining task Masked Topology Modeling (MTM) that leverages the face-adjacency graph, a graph whose edge attributes the encoder can be asked to predict. MTM masks a fraction of edges and trains a small head to predict each masked edge's convexity and curve type from the encoder's post-message-passing face features. We combine MTM with a MoCo-style momentum-queue contrastive learning over B-rep-aware augmentations and pretrain on the ABC dataset and our new procedurally generated dataset to show strong performance on a number of benchmarks. We further demonstrate MTM's effectiveness across B-rep architectures.

Primary area:
unsupervised, self-supervised, semi-supervised, and supervised representation learning
Keywords:
self-supervised learning, boundary representation, computer-aided design

Reviews

Among:
Shape:
1234this paper, Bottom 10%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 1.67 over 3 reviews, above 10% of the others.
Reviewer_AqY1OpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
3
Soundness
2
Read review
Paper summary

The paper proposes Masked Topology Modeling (MTM), a self-supervised learning approach for learning representations of parametric CAD models in B-rep format. The method represents a B-rep as a face-adjacency graph, masks some of the connections between faces, and trains an encoder to predict properties of the masked edges, including their convexity and curve type. Since these properties can be obtained directly from the B-rep, the model can be pretrained without manually labeled data. The authors combine MTM with contrastive learning and evaluate the learned representation on several downstream CAD classification and segmentation tasks. The experiments show that MTM is particularly effective when only a small amount of labeled data is available, while its advantage becomes smaller when sufficient labeled training data is used. Overall, the paper demonstrates that the intrinsic topology of B-rep models can be used as a self-supervised signal to improve the label efficiency of CAD representation learning.

Critical strengths

The paper addresses an interesting and important problem of learning useful representations for CAD models when labeled data is limited. The observation that masking-based self-supervised learning, which has been successful in language and image domains, can also be effectively applied to the topology of B-rep models is interesting. In particular, predicting properties of masked edges provides a natural way of using the information already available in CAD models as supervision. Another strength is the extensiveness of the experiments. The authors evaluate the approach on multiple datasets, downstream tasks, architectures, and different amounts of labeled data, and provide several ablation studies to understand the contribution of different components of the method.

Critical weaknesses

My main concern is the practicality of the proposed method. The largest improvements are observed in extremely low-data settings, such as using only 0.1% of the labeled data. While these experiments demonstrate the label efficiency of the learned representation, it is unclear how representative such settings are of practical CAD applications, where labeled B-rep datasets are available and additional synthetic data can also be generated. When more labeled data is available, the advantage of the proposed method becomes considerably smaller and in some experiments it does not outperform existing approaches. This raises the question of whether the additional pretraining stage provides a significant practical benefit outside very limited-data scenarios.

Questions for authors

The largest improvements seem to happen in extremely low-data settings, such as 0.1% labeled data. What practical CAD applications do the authors envision where labeled data is this scarce?

When more labeled data is available, the improvement from MTM becomes relatively small and sometimes disappears. How do the authors justify the additional pretraining cost in these settings?

Given that synthetic B-rep data can be procedurally generated, why is self-supervised pretraining preferable to simply generating more labeled or automatically annotated training data?

Masking-based representation learning has been extensively explored in language and vision. Beyond adapting masking to B-rep topology, what do the authors consider the main conceptual novelty of MTM?

The method predicts convexity and curve type for masked edges. Why were these two properties selected?

Reviewer_XVdUOpenReview ↗
Rating
1
Confidence
4
Informativeness
2
Presentation
2
Soundness
1
Read review
Paper summary

This paper proposes Masked Topology Modeling, a self-supervised learning objective for learning representations of parametric CAD models in their native B-Rep format. MTM operates on face-adjacency graphs by masking selected edges and predicting their convexity and curve types from the contextual representations of adjacent faces. The authors combine MTM with contrastive learning and B-Rep-specific augmentations, using both the ABC dataset and a newly generated synthetic CAD dataset for pretraining. The paper also provides theoretical analyses of the identifiability and separability of the masked edge attributes.

Critical strengths

The MTM setup is insightful.

The b-rep question and motivation is interesting.

Critical weaknesses

This work is clearly AI generated. While the community may still be under debate of whether ai based auto research works should be accepted, I personally question the efforts being put into this particular work and hence cannot accept this version. In addition, two points to the quality of the content:

  1. Although the paper proposes a self-supervised representation learning objective, its evaluation spans several different downstream tasks with inconsistent metrics and experimental protocols. The presentation makes it difficult to understand the primary objective and assess the practical significance of the reported improvements.
  2. Many primary baselines are relatively old, while recent self-supervised approaches such as BRepMAE and HierMAE are excluded from several comparisons.
Questions for authors

See weaknesses.

Reviewer_dgCDOpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

This paper proposes Masked Topology Modeling (MTM), a self-supervised pretraining method for B-rep CAD representations. It masks a subset of edges in the face-adjacency graph and predicts their convexity and curve-type attributes using face geometry and the remaining graph context. The method is combined with contrastive learning and evaluated on downstream tasks including face segmentation, shape classification, and instance segmentation. The authors report improvements over several baselines, particularly in low-label settings.

Critical strengths

The paper addresses a meaningful problem: labeled data are scarce for downstream CAD tasks, and self-supervised pretraining can exploit the geometric and topological information already available in B-rep models without requiring additional manual annotations.

Critical weaknesses

1. The motivation for the specific masking objective is insufficiently established. Transferring masked modeling to CAD is reasonable, but B-rep already provides structured face geometry, bounded surface regions, and known face-adjacency relationships. MTM preserves the identities and geometric features of adjacent faces and masks only the corresponding edge connection during message passing, while predicting its convexity and curve type. These attributes may be strongly constrained by the local geometry of the incident faces. It is therefore unclear whether this task requires learning meaningful topological context, rather than primarily learning to predict local geometric attributes.

2. The contribution of MTM beyond simpler pretraining objectives remains unclear. The downstream results suggest that MTM can be useful, particularly when labeled data are limited. However, the paper should more clearly establish whether its specific edge-masking objective provides meaningful advantages over simpler geometric prediction or graph-based self-supervised objectives using the same CAD data.

Questions for authors

see weakness

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Loading the map…

Discussion (0)

Sign in to comment.