Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Retrieval-Augmented Multi-LLM Ensemble for Industrial Part Specification Extraction

Muzakkiruddin Ahmed Mohammed, John R. Talburt, Leon Claasssens, Adriaan Marais

TL;DR

The paper tackles the challenge of extracting precise industrial part specifications from unstructured text, a task hindered by hallucinations and uneven model performance. It introduces RAGsemble, a nine-LLM ensemble connected through a three-phase, retrieval-augmented pipeline that grounds outputs in FAISS-based knowledge retrieval and a central synthesis layer. Key contributions include scalable ensemble architecture, seamless RAG grounding, transparent confidence and consensus mechanisms, and a deployable implementation for manufacturing domains. Results on real industrial data, including GE9X, show substantial improvements in completeness, technical depth, and formatting reliability over strong single-model baselines, demonstrating practical potential for production deployment.

Abstract

Industrial part specification extraction from unstructured text remains a persistent challenge in manufacturing, procurement, and maintenance, where manual processing is both time-consuming and error-prone. This paper introduces a retrieval-augmented multi-LLM ensemble framework that orchestrates nine state-of-the-art Large Language Models (LLMs) within a structured three-phase pipeline. RAGsemble addresses key limitations of single-model systems by combining the complementary strengths of model families including Gemini (2.0, 2.5, 1.5), OpenAI (GPT-4o, o4-mini), Mistral Large, and Gemma (1B, 4B, 3n-e4b), while grounding outputs in factual data using FAISS-based semantic retrieval. The system architecture consists of three stages: (1) parallel extraction by diverse LLMs, (2) targeted research augmentation leveraging high-performing models, and (3) intelligent synthesis with conflict resolution and confidence-aware scoring. RAG integration provides real-time access to structured part databases, enabling the system to validate, refine, and enrich outputs through similarity-based reference retrieval. Experimental results using real industrial datasets demonstrate significant gains in extraction accuracy, technical completeness, and structured output quality compared to leading single-LLM baselines. Key contributions include a scalable ensemble architecture for industrial domains, seamless RAG integration throughout the pipeline, comprehensive quality assessment mechanisms, and a production-ready solution suitable for deployment in knowledge-intensive manufacturing environments.

Retrieval-Augmented Multi-LLM Ensemble for Industrial Part Specification Extraction

TL;DR

The paper tackles the challenge of extracting precise industrial part specifications from unstructured text, a task hindered by hallucinations and uneven model performance. It introduces RAGsemble, a nine-LLM ensemble connected through a three-phase, retrieval-augmented pipeline that grounds outputs in FAISS-based knowledge retrieval and a central synthesis layer. Key contributions include scalable ensemble architecture, seamless RAG grounding, transparent confidence and consensus mechanisms, and a deployable implementation for manufacturing domains. Results on real industrial data, including GE9X, show substantial improvements in completeness, technical depth, and formatting reliability over strong single-model baselines, demonstrating practical potential for production deployment.

Abstract

Industrial part specification extraction from unstructured text remains a persistent challenge in manufacturing, procurement, and maintenance, where manual processing is both time-consuming and error-prone. This paper introduces a retrieval-augmented multi-LLM ensemble framework that orchestrates nine state-of-the-art Large Language Models (LLMs) within a structured three-phase pipeline. RAGsemble addresses key limitations of single-model systems by combining the complementary strengths of model families including Gemini (2.0, 2.5, 1.5), OpenAI (GPT-4o, o4-mini), Mistral Large, and Gemma (1B, 4B, 3n-e4b), while grounding outputs in factual data using FAISS-based semantic retrieval. The system architecture consists of three stages: (1) parallel extraction by diverse LLMs, (2) targeted research augmentation leveraging high-performing models, and (3) intelligent synthesis with conflict resolution and confidence-aware scoring. RAG integration provides real-time access to structured part databases, enabling the system to validate, refine, and enrich outputs through similarity-based reference retrieval. Experimental results using real industrial datasets demonstrate significant gains in extraction accuracy, technical completeness, and structured output quality compared to leading single-LLM baselines. Key contributions include a scalable ensemble architecture for industrial domains, seamless RAG integration throughout the pipeline, comprehensive quality assessment mechanisms, and a production-ready solution suitable for deployment in knowledge-intensive manufacturing environments.
Paper Structure (8 sections, 5 equations, 2 figures, 1 table)

This paper contains 8 sections, 5 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Literature Review
  3. System Architecture
  4. Methodology and Implementation
  5. Results and Analysis
  6. Use Cases and Applications
  7. Limitations and Future Work
  8. Conclusions

Figures (2)

  • Figure 1: System architecture of the multi-LLM industrial part extraction framework, showing modular components and model organization.
  • Figure 2: Three-phase multi-LLM processing workflow with continuous RAG integration. Each stage contributes uniquely to extraction, enrichment, and validation of part specifications.