![[2509.01350] Error Notebook-Guided, Training-Free Part Retrieval in 3D CAD Assemblies via Vision-Language Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2509.01350] Error Notebook-Guided, Training-Free Part Retrieval in 3D CAD Assemblies via Vision-Language Models
Summary
The paper presents a novel framework for part retrieval in 3D CAD assemblies using vision-language models, emphasizing training-free methods and improved accuracy through Error Notebooks.
Why It Matters
This research addresses the challenges of using large language models for CAD part retrieval, particularly in handling complex metadata without extensive training. The proposed method enhances retrieval accuracy, making it valuable for automated engineering tasks and advancing the application of AI in design.
Key Takeaways
- Introduces a two-stage framework for part retrieval in CAD assemblies.
- Utilizes Error Notebooks to improve retrieval accuracy without training.
- Demonstrates significant performance gains over existing training-free methods.
Computer Science > Artificial Intelligence arXiv:2509.01350 (cs) [Submitted on 1 Sep 2025 (v1), last revised 25 Feb 2026 (this version, v3)] Title:Error Notebook-Guided, Training-Free Part Retrieval in 3D CAD Assemblies via Vision-Language Models Authors:Yunqing Liu, Nan Zhang, Zhiming Tan View a PDF of the paper titled Error Notebook-Guided, Training-Free Part Retrieval in 3D CAD Assemblies via Vision-Language Models, by Yunqing Liu and 2 other authors View PDF HTML (experimental) Abstract:Effective specification-aware part retrieval within complex CAD assemblies is essential for automated engineering tasks. However, using LLMs/VLMs for this task is challenging: the CAD model metadata sequences often exceed token budgets, and fine-tuning high-performing proprietary models (e.g., GPT or Gemini) is unavailable. Therefore, we need a framework that delivers engineering value by handling long, non-natural-language CAD model metadata using VLMs, but without training. We propose a 2-stage framework with inference-time adaptation that combines corrected Error Notebooks with RAG to substantially improve VLM-based part retrieval reasoning. Each Error Notebook is built by correcting initial CoTs through reflective refinement, and then filtering each trajectory using our proposed grammar-constraint (GC) verifier to ensure structural well-formedness. The resulting notebook forms a high-quality repository of specification-CoT-answer triplets, from which RAG retrieves specification-rele...
