CADReview

KO | EN

gitlite — search

#python #hacktoberfest #typescript #go #golang #rust #deep-learning #linux #nodejs #kubernetes #vue #pytorch #ios #openai

(★ 23)

[ACL 2025 Oral] The official repository of our paper: CADReview: Automatically Reviewing CAD Programs with Error Detection and Correction

CADReview 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# CADReview :building_construction:
[![arXiv](https://img.shields.io/badge/arXiv-2505.22304-b31b1b.svg)](https://arxiv.org/pdf/2505.22304)

[**🌐 Homepage**](https://cgl-pro.github.io/cadreview/)  | [**📖 arXiv**](https://arxiv.org/abs/2505.22304) | [**🤗 HuggingFace Dataset**](https://huggingface.co/datasets/Garygedegege/CADReview)

This repo contains the codebase for our paper [CADReview: Automatically Reviewing CAD Programs with Error Detection and Correction](https://arxiv.org/abs/xxxx)

**🎉 ACL 2025 main**

## Introduction

We introduce the **CAD review** task, which aims to automatically detect and correct errors in CAD programs by comparing them with reference images. To support this task, we propose **ReCAD**, a multimodal large language model (MLLM)-based framework that generates feedback and edits code for accurate 3D reconstruction. We also present **CADReview**, a large-scale dataset with over 20,000 program–image pairs featuring diverse geometric structures and real-world error types. Our results show that ReCAD significantly outperforms existing models, offering a practical solution for AI-assisted CAD debugging and refinement.

![](https://raw.githubusercontent.com/Gary-code/CADReview/main/images/com_pic.png)

## Training and Inference

1. Our training and inference are conducted using the [ms-swift](https://github.com/modelscope/ms-swift) framework. Environment configuration: `ms-swift >= 3.3`, `vllm >= 0.7.3`.
2. The alignment training for GCR and SGO can be found in:
   `./training_and_inference/alignment_gcr` and `./training_and_inference/alignment_sgo`.
3. Training for $\phi_1$ and $\phi_2$ can be found in:
   `./training_and_inference/feedback_gen` and `./training_and_inference/code_editor`.
4. The inference script can be found at:
   `./training_and_inference/inference.py`.

## Evaluation

Run `./evaluate/eval.sh` to perform evaluation.

CADReview

CADReview 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

CADReview

CADReview 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README