HistCAD: A Constraint-Aware Parametric History-Based CAD Representation, Dataset, and Benchmark with Industrial Complexity
HistCAD introduces a constraint-aware, history-based CAD representation meant to be executable across CAD systems, not just visually similar. It records sketch primitives, constraints, feature operations, and 3D boundary references for operations like fillets and chamfers, so edits can be tested against the original design logic instead of only the final mesh.
The dataset is large enough to matter: 170,236 executable sequences paired with native CAD models, STEP files, rendered views, and text annotations. The benchmark adds three edit-focused metrics—Edit Reachability (ER), conditional preserved constraint satisfaction (cPCSR), and Overall Editable Success (OES)—to separate “can the edit be reached?” from “does the result still obey the constraints?” That distinction is useful for anyone working on CAD generation, text-to-CAD, or LLM-assisted design workflows, because a pretty output that collapses under the first parameter tweak is not very useful in practice.
“Existing CAD generation datasets and benchmarks emphasize reconstruction fidelity... leaving preservation of design intent under edits largely...”
- what
- HistCAD is a constraint-aware, parametric history-based CAD representation, dataset, and benchmark focused on editability.
- who
- Paper by Xintong Dong, Chuanyang Li, Peng Zheng, Chuqi Han, Jiaxin Jing, Hailong Shen, Yanzhi Song, and Zhouwang Yang.
- when
- arXiv v1 submitted Dec. 8, 2025; v3 revised May 29, 2026.
- impact
- It gives CAD generation and LLM-based design tools a way to measure whether generated models remain valid after parameter edits, not just whether they look right.
Useful benchmark that measures real editability, not just shape quality
Follow parametric modeling updates
See relevant stories in your personalized news feed.
Discussion