Bridging Industrial Expertise and XR with LLM-Powered Conversational Agents

Overview

This paper (arXiv:2504.05527, April 2025) investigates how LLM-powered conversational agents can bridge industrial expertise and extended reality (XR). It belongs to the Maintenance, Repair, Manufacturing domain of this collection.

• Authors: Despina Tomkou, George Fatouros, Andreas Andreou, Georgios Makridis, Fotis Liarokapis, Dimitrios Dardanis, et al. (9 authors in total)
• Source: https://arxiv.org/abs/2504.05527
• Type: Academic paper

Motivation

Industrial domains such as maintenance, repair, and manufacturing depend heavily on scarce expert knowledge. At the same time, XR (VR/AR/MR) training and assistance systems need ways to deliver that knowledge contextually inside immersive environments. This work proposes LLM-based conversational agents as the connecting layer: workers in XR scenarios can interact in natural language with an agent that draws on industrial knowledge sources, reducing the dependency on physically available experts.

Core Contribution

• A unified approach that combines conversational LLM agents with XR interfaces for industrial knowledge delivery.
• Integration of retrieval/generation components with domain (industrial) knowledge sources — an instance of domain-specific agentic RAG.
• Discussion of the path from research prototypes to deployable industrial systems, including constraints such as latency, cost, hallucination control, and safety.

Relevance to Search / RAG Systems

From a systems perspective, the paper addresses a core question of the LLM era: how to allocate responsibilities across retrieval, ranking, generation, and tool invocation. Typical considerations for such pipelines include:

1. Representation — encoding queries, documents, and user context as dense/sparse representations or structured prompts; 2. Core modules — retrievers, re-rankers, planners, memory, and tool interfaces; 3. Learning/inference strategies — iterative retrieval, sub-query decomposition, early stopping, and budget control.

Open Problems

• Evaluation credibility for agentic, task-oriented retrieval (task success rate, citation accuracy rather than static nDCG alone);
• Latency, token cost, and real-time constraints in XR/industrial settings;
• Hallucination and safety risks when agents operate on industrial knowledge;
• Cross-lingual and multimodal generalization.

Related Entries

• A Compliance-Preserving Retrieval System for Aircraft MRO Task Search
• Enhancing Manufacturing Knowledge Access with LLMs and Context-aware Prompting
• MetalMind: knowledge graph-driven human-centric knowledge system
• Optimizing Aerospace Product Maintenance: Multi-Modal Knowledge approach
• Prescriptive Agents based on RAG for Automated Maintenance (PARAM)

> Note: Quantitative results should be verified against the original PDF; this entry is compiled from the abstract and public metadata.

References

• Original paper: "Bridging Industrial Expertise and XR with LLM-Powered Conversational Agents", arXiv:2504.05527, April 2025. https://arxiv.org/abs/2504.05527