Search papers, labs, and topics across Lattice.
BAIRBeihangDeepSeekHKUShanghaiTechShenzhen Loop Area InstituteTencent AITranscengramMar 4, 2026arXiv:2603.04337Pointer-CAD is introduced as a novel LLM-based framework for CAD model generation that uses pointer-based command sequences to incorporate B-rep geometric information into sequential modeling. It addresses limitations of prior command sequence methods by enabling explicit entity selection (faces/edges) and reducing quantization errors via pointers to geometric entities. Experiments on a newly created dataset of 575K CAD models demonstrate Pointer-CAD's ability to generate complex structures and significantly reduce segmentation errors compared to existing approaches.
LLMs can now generate more accurate and complex CAD models by pointing to existing geometric entities, rather than relying on discretized command sequences prone to topological errors.
Constructing computer-aided design (CAD) models is labor-intensive but essential for engineering and manufacturing. Recent advances in Large Language Models (LLMs) have inspired the LLM-based CAD generation by representing CAD as command sequences. But these methods struggle in practical scenarios because command sequence representation does not support entity selection (e.g. faces or edges), limiting its ability to support complex editing operations such as chamfer or fillet. Further, the discretization of a continuous variable during sketch and extrude operations may result in topological errors. To address these limitations, we present Pointer-CAD, a novel LLM-based CAD generation framework that leverages a pointer-based command sequence representation to explicitly incorporate the geometric information of B-rep models into sequential modeling. In particular, Pointer-CAD decomposes CAD model generation into steps, conditioning the generation of each subsequent step on both the textual description and the B-rep generated from previous steps. Whenever an operation requires the selection of a specific geometric entity, the LLM predicts a Pointer that selects the most feature-consistent candidate from the available set. Such a selection operation also reduces the quantization error in the command sequence-based representation. To support the training of Pointer-CAD, we develop a data annotation pipeline that produces expert-level natural language descriptions and apply it to build a dataset of approximately 575K CAD models. Extensive experimental results demonstrate that Pointer-CAD effectively supports the generation of complex geometric structures and reduces segmentation error to an extremely low level, achieving a significant improvement over prior command sequence methods, thereby significantly mitigating the topological inaccuracies introduced by quantization error.
