Ziheng Zhang

Paper [15], in preparation

VeriBIM: Natural-Language Editing of IFC Building Models with a Verifier-Trained, Locally Deployable Language Model

Zhang, Z., and Zhang, W.*

Automation in Construction (target venue). In preparation.

Summary

VeriBIM is a locally deployable language model that edits IFC building models from natural-language instructions. An IFC verifier that scores geometry, semantics and topology is used to filter training data, to label preference pairs, and as the reward in reinforcement learning.

Cite

@unpublished{zhangveribim,
  title     = {VeriBIM: Natural-Language Editing of IFC Building Models with a Verifier-Trained, Locally Deployable Language Model},
  author    = {Zhang, Z. and Zhang, W.},
  note      = {In preparation, target venue Automation in Construction}
}

All publications (15)

  1. [11]

    CertiGATE: A Guard Benchmark and Quality-Certificate Protocol for Large Language Model-Based Work-Order Scheduling in Facility Management

    Zhang, Z.*, Ku Chia, T. P., Yang, X., Chang, X., Wang, S., and Zhang, W.*

    Advanced Engineering Informatics. Overview figure