CADBench (Doris et al.)

CADBench is a unified benchmark for multimodal CAD program generation, built because existing evaluations are fragmented across datasets, modalities and metrics. It contains 18,000 evaluation samples spanning six benchmark families derived from DeepCAD, Fusion 360, ABC, MCB and Objaverse, with five input modalities including clean meshes, noisy meshes, single-view renders, photorealistic renders and multi-view renders.

It scores generated CAD programs on six metrics covering geometric fidelity, executability and program compactness, including volumetric IoU against the reference. STEP-based families are stratified by B-rep face count and all families are diversity-sampled to support controlled analysis across complexity and object variation.

The paper benchmarks eleven CAD-specialized and general-purpose vision-language systems, generating more than 1.4 million CAD programs. Under idealized inputs, specialized mesh-to-CAD models substantially outperform code-generating VLMs, which remain far from reliable CAD program reconstruction. The paper reports three recurring failure modes: quality degrades with geometric complexity, CAD-specialized models can be brittle under modality shift, and model rankings change across metrics.