ARTFEED — Contemporary Art Intelligence

CADIR: New Intermediate Representation for Cross-Backend CAD Generation

ai-technology · 2026-08-04

A team of researchers has unveiled CADIR, an executable intermediate representation designed for agent-friendly computer-aided design (CAD) generation and cross-backend editing. Utilizing the OCCT geometry kernel through OCP, CADIR offers explicit and compositional modeling operations along with detailed execution diagnostics. Throughout the execution of a program, CADIR captures modeling operations, parameter dependencies, constraints, and topology choices in a construction graph. To facilitate dependable cross-backend reconstruction, the approach incorporates Geometric Signature Matching. This research tackles the shortcomings of current methods that depict modeling processes as backend-specific sequential scripts or static geometry, which impede the preservation of construction history, stable topological references, and feature-level editability across various CAD systems. The paper can be found on arXiv with the identifier 2608.00891.

Key facts

  • CADIR is a new intermediate representation for CAD generation and editing.
  • It is built on the OCCT geometry kernel via OCP.
  • CADIR provides explicit, compositional modeling operations and fine-grained execution diagnostics.
  • It records modeling operations, parameter dependencies, constraints, and topology selections in a construction graph.
  • Geometric Signature Matching is introduced for reliable cross-backend reconstruction.
  • The method aims to preserve construction history, stable topological references, and feature-level editability across CAD systems.
  • The paper is available on arXiv with identifier 2608.00891.
  • The work addresses limitations of backend-specific sequential scripts and static geometry.

Entities

Sources