Reproduce “CReFT-CAD: Boosting Orthographic Projection Reasoning for CAD via Reinforcement Fine-Tuning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ke Niu, Zhuofan Chen, Haiyang Yu, Yuwen Chen, Teng Fu, Mengyang Zhao, Bin Li, Xiangyang Xue
CReFT-CAD trains a vision-language model to interpret orthographic drawings through staged checking, parameter selection, and calculation, followed by supervised post-tuning. On 3,000 real-world prefabricated bridge-pier drawings, it reaches 46.67% aggregate parameter accuracy, compared with 36.15% for standard SFT and 13.47% for Qwen2.5-VL. These scores count parameters, not fully correct drawings, and do not isolate reinforcement learning’s effect.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.