OpenNovelty.org

Auto-Regressive Surface Cutting

ICLR 2026 Conference SubmissionAnonymous Authors
OpenReview Score: 6.5Download Report PDF
surface cuttingMesh UV unfolding3D part segmentationauto-regressive generative model
Abstract:

Surface cutting is a fundamental task in computer graphics, with applications in UV parameterization, texture mapping, and mesh decomposition. However, existing methods often produce technically valid but overly fragmented atlases that lack semantic coherence. We introduce SeamGPT, an auto-regressive model that generates cutting seams by mimicking professional workflows. Our key technical innovation lies in formulating surface cutting as a next token prediction task: sample point clouds on mesh vertices and edges, encode them as shape conditions, and employ a GPT-style transformer to sequentially predict seam segments with quantized 3D coordinates. Our approach achieves exceptional performance on UV unwrapping benchmarks containing both manifold and non-manifold meshes, including artist-created, and 3D-scanned models. In addition, it enhances existing 3D segmentation tools by providing clean boundaries for part decomposition.

Disclaimer
This report is AI-GENERATED using Large Language Models and WisPaper (A scholar search engine). It analyzes academic papers' tasks and contributions against retrieved prior work. While this system identifies POTENTIAL overlaps and novel directions, ITS COVERAGE IS NOT EXHAUSTIVE AND JUDGMENTS ARE APPROXIMATE. These results are intended to assist human reviewers and SHOULD NOT be relied upon as a definitive verdict on novelty.
NOTE that some papers exist in multiple, slightly different versions (e.g., with different titles or URLs). The system may retrieve several versions of the same underlying work. The current automated pipeline does not reliably align or distinguish these cases, so human reviewers will need to disambiguate them manually.
If you have any questions, please contact: mingzhang23@m.fudan.edu.cn

Overview

Taxonomy

Core-task Taxonomy Papers
3
3
Claimed Contributions
17
Contribution Candidate Papers Compared
0
Refutable Paper

Research Landscape Overview

Core task: auto-regressive surface cutting for mesh UV parameterization. The field of mesh UV parameterization has evolved along several complementary directions, each addressing different aspects of the challenge of flattening 3D surfaces onto 2D texture space. Learning-Based Sequential Seam Generation explores data-driven approaches that predict where to cut meshes in a step-by-step fashion, often leveraging neural architectures to learn optimal seam placement from examples. Joint Optimization of Cuts and Parameterization seeks to simultaneously determine both the cut locations and the resulting UV layout, treating these traditionally separate stages as a unified problem; Optcuts[1] exemplifies this approach by co-optimizing distortion and seam quality. Part-Based and Semantic UV Unwrapping focuses on decomposing meshes according to meaningful geometric or semantic boundaries, as seen in PartUV[2], which leverages part-aware strategies to produce more intuitive texture atlases. Progressive Mesh Parameterization, illustrated by Texture Mapping Progressive Meshes[3], addresses parameterization in the context of level-of-detail hierarchies, ensuring that UV coordinates remain consistent across mesh simplifications. A central tension across these branches involves the trade-off between automation and control: fully automatic methods may struggle with artist-desired layouts, while semantic or part-based approaches require additional annotations or segmentation. Another active theme is the balance between local geometric quality (minimizing distortion) and global atlas efficiency (reducing seam length and chart count). Auto-Regressive Surface Cutting[0] sits within the Learning-Based Sequential Seam Generation branch, specifically employing transformer-based architectures to predict seam edges in an auto-regressive manner. This positions it as a data-driven alternative to joint optimization methods like Optcuts[1], emphasizing learned sequential decision-making over explicit energy minimization, and contrasting with part-based strategies such as PartUV[2] by focusing on direct seam prediction rather than semantic decomposition.

Claimed Contributions

SeamGPT: Auto-regressive model for surface cutting seam generation

The authors propose SeamGPT, which formulates surface cutting as a next-token prediction task using a GPT-style transformer to sequentially predict seam segments with quantized 3D coordinates, mimicking artist workflows for generating semantically meaningful cutting seams.

6 retrieved papers
Auto-regressive formulation of cutting seams as line segment sequences

The authors introduce a novel formulation that represents cutting seams as ordered sequences of 3D line segments with quantized coordinates, enabling auto-regressive generation through a GPT-inspired transformer architecture that captures sequential dependencies in the cutting process.

1 retrieved paper
Seam-enhanced 3D part segmentation approach

The authors demonstrate that SeamGPT-generated seams can enhance existing 3D part segmentation methods by providing clean part boundaries, addressing the limitation of blurry boundaries in current segmentation tools through a patch-based part segmentation methodology.

10 retrieved papers

Core Task Comparisons

Comparisons with papers in the same taxonomy category

Within the taxonomy built over the current TopK core-task papers, the original paper is assigned to a leaf with no direct siblings and no cousin branches under the same grandparent topic. In this retrieved landscape, it appears structurally isolated, which is one partial signal of novelty, but still constrained by search coverage and taxonomy granularity.

Contribution Analysis

Detailed comparisons for each claimed contribution

Contribution

SeamGPT: Auto-regressive model for surface cutting seam generation

The authors propose SeamGPT, which formulates surface cutting as a next-token prediction task using a GPT-style transformer to sequentially predict seam segments with quantized 3D coordinates, mimicking artist workflows for generating semantically meaningful cutting seams.

Contribution

Auto-regressive formulation of cutting seams as line segment sequences

The authors introduce a novel formulation that represents cutting seams as ordered sequences of 3D line segments with quantized coordinates, enabling auto-regressive generation through a GPT-inspired transformer architecture that captures sequential dependencies in the cutting process.

Contribution

Seam-enhanced 3D part segmentation approach

The authors demonstrate that SeamGPT-generated seams can enhance existing 3D part segmentation methods by providing clean part boundaries, addressing the limitation of blurry boundaries in current segmentation tools through a patch-based part segmentation methodology.

Table of Contents

Auto-Regressive Surface Cutting | Novelty Validation