LLM Adaptation for Polymer-Composite Additive Manufacturing via RAG and Fine-Tuning
A study from arXiv (2605.12516) explores domain adaptation of large language models (LLMs) for polymer-composite additive manufacturing (AM). General-purpose LLMs struggle with specialized engineering domains due to limited domain grounding. The research investigates fine-tuning and retrieval-augmented generation (RAG) to improve answer accuracy and relevance. A curated AM corpus was constructed, and three configurations based on LLaMA-3 were evaluated. The work addresses the challenge of heterogeneous AM knowledge sources, including academic literature, manufacturer documentation, technical standards, and procedural guides.
Key facts
- arXiv paper 2605.12516
- Domain adaptation of LLMs for polymer-composite additive manufacturing
- Uses retrieval-augmented generation (RAG) and fine-tuning
- Constructs a curated AM corpus
- Evaluates three configurations based on LLaMA-3
- Addresses heterogeneous knowledge sources: academic literature, manufacturer documentation, technical standards, procedural guides
- General-purpose LLMs lack domain grounding for specialized engineering
- Focus on improving answer accuracy, relevance, and usability for expert-level QA
Entities
Institutions
- arXiv
- LLaMA-3