Llms predict and explain materials for vehicle parts reliably
Exploring LLMs and RAG for Plausible and Explainable Material Prediction of Vehicle Components
Information Retrieval
Summary
The paper looks at whether large language models (LLMs), like AI that can write and answer questions, can figure out what materials are used for car parts such as brake discs or fuel injectors. The authors tested three methods, including one where the AI gets help by looking up information from a special Wikipedia collection. They found that just using the AI alone works well, better than previous methods, and the extra information lookup did not improve results. They also built a tool to help experts check how good the AI's answers are. Some challenges remain, like finding better data and tuning the system to work better.
What this means in practice
- •For automotive engineers: Suggest plausible materials for vehicle components during design without needing extensive data or retraining.
- •For technical writers: Generate explainable material descriptions for vehicle parts based on existing public knowledge to assist documentation.
Authors
Frederik Wagner, Annerose Eichel, Sabine Schulte im Walde
Abstract
In this work, we explore whether LLMs can accurately predict and explain plausible materials for vehicle components such as brake discs or fuel injectors without requiring extensive fine-tuning. We test and evaluate three approaches: a standard generative LLM baseline, a single-pass Retrieval-Augmented Generation (RAG) approach, and an iterative Chain-of-Verification (CoVe) variant. For retrieval, we rely on publicly available data using a domain-filtered Wikipedia corpus. Since no gold standard exists for this task, we develop a custom web-based annotation tool supporting crucial functions for structured domain expert evaluation. LLM-based generation substantially outperforms prior work, which is not further surpassed by the tested RAG approaches. Our results surface remaining challenges for RAG-based systems: hyperparameter optimization, the availability of high-quality, legally accessible domain corpora, and expert evaluation study design.