Research map / Probabilistic and scientific modeling / Molecules and proteins
AutoRegressive Generation with B-rep Holistic Token Sequence Representation
Jiahao Li, Yunpeng Bai, Yongkang Dai, Hao Guo, Hongping Gan, Yilei Shi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add DFS face ordering instead of random order (ablation) | BrepARG · DeepCAD unconditional | Validity (%): 67.92 -> 87.60 | Improved |
| add DFS face ordering instead of BFS (ablation) | BrepARG · DeepCAD unconditional | Validity (%): 85.94 -> 87.60 | Improved |
| add DFS face ordering instead of ZYX centroid ordering (ablation) | BrepARG · DeepCAD unconditional | Validity (%): 83.12 -> 87.60 | Improved |
| add DFS face ordering instead of spectral ordering (ablation) | BrepARG · DeepCAD unconditional | Validity (%): 79.38 -> 87.60 | Improved |
| add MAX-IDX-A edge ordering instead of random (ablation) | BrepARG · DeepCAD unconditional | Validity (%): 85.43 -> 87.60 | Improved |
| add MAX-IDX-A edge ordering instead of random (ablation) | BrepARG · DeepCAD unconditional | COV (%): 74.25 -> 75.45 | Improved |
| increase VQ-VAE codebook size from 512 to 4096 (ablation) | Face-Edge VQ-VAE · DeepCAD unconditional | Validity (%): 69.8% -> 87.5% | Improved |
| lower nucleus sampling p from 0.95 to 0.6 (ablation) | BrepARG · DeepCAD unconditional | Validity (%): 82.85 -> 90.25 | Mixed |
| lower nucleus sampling p from 0.9 to 0.6 (ablation) | BrepARG · DeepCAD unconditional | COV (%): 75.45 -> 63.50 | Hurt |
| use deep-learning quantization of bounding boxes (best scheme 2) instead of uniform scalar quantization (ablation) | position tokenization · bounding box reconstruction | Acc: 99.4% | Hurt |
| use BrepARG instead of DTGBrepGen | DTGBrepGen · DeepCAD unconditional | Validity (%): 79.80 -> 87.60 | Improved |
| use BrepARG instead of DTGBrepGen | DTGBrepGen · ABC unconditional | Validity (%): 57.59 -> 67.54 | Improved |
Closest papers by meaning
- DTGBrepGen: A Novel B-rep Generative Model through Decoupling Topology and Geometry (CVPR 2025 · similarity 0.86)
- BrepGiff: Lightweight Generation of Complex B-rep with 3D GAT Diffusion (CVPR 2025 · similarity 0.85)
- Hierarchical Masked Autoregressive Models with Low-Resolution Token Pivots (ICML 2025 · similarity 0.84)
- BitDance: Scaling Autoregressive Generative Models with Binary Tokens (NeurIPS 2026 · similarity 0.82)
- Vision Foundation Models as Effective Visual Tokenizers for Autoregressive Generation (NeurIPS 2025 · similarity 0.82)
- ImageFolder: Autoregressive Image Generation with Folded Tokens (ICLR 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
