TeleMoM: Consensus-Driven Telecom Intelligence via Mixture of Models

2025-04-03

AI Summary

Large language models (LLMs) face significant challenges in telecommunications—characterized by dense domain-specific terminology, rapid knowledge evolution, and high technical complexity—limiting their practical applicability. Method: This paper proposes a consensus-driven multi-LLM collaboration framework featuring a novel two-stage consensus mechanism: (i) multiple proposer LLMs generate rationale-augmented responses; (ii) an adjudicator LLM dynamically evaluates response quality and synthesizes final outputs via domain-adapted consensus aggregation. The framework integrates multi-LLM collaborative reasoning, a proposer-adjudicator architecture, and adaptive consensus strategies. Contribution/Results: It overcomes limitations of single-model and naive ensemble approaches by jointly suppressing bias and enabling cross-model knowledge complementarity. Evaluated on telecom-specific tasks, the framework achieves a 9.7% absolute accuracy gain over strong baselines—including RAG, Mixture-of-Experts (MoE), and fine-tuning—demonstrating superior scalability and robustness. It establishes a new paradigm for deploying LLMs in fast-evolving, knowledge-intensive professional domains.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Multiagent Systems: Multiagent Learning

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Large language models (LLMs) face significant challenges in specialized domains like telecommunication (Telecom) due to technical complexity, specialized terminology, and rapidly evolving knowledge. Traditional methods, such as scaling model parameters or retraining on domain-specific corpora, are computationally expensive and yield diminishing returns, while existing approaches like retrieval-augmented generation, mixture of experts, and fine-tuning struggle with accuracy, efficiency, and coordination. To address this issue, we propose Telecom mixture of models (TeleMoM), a consensus-driven ensemble framework that integrates multiple LLMs for enhanced decision-making in Telecom. TeleMoM employs a two-stage process: proponent models generate justified responses, and an adjudicator finalizes decisions, supported by a quality-checking mechanism. This approach leverages strengths of diverse models to improve accuracy, reduce biases, and handle domain-specific complexities effectively. Evaluation results demonstrate that TeleMoM achieves a 9.7% increase in answer accuracy, highlighting its effectiveness in Telecom applications.

Problem

• LLMs struggle in Telecom due to complexity and terminology
• Existing methods lack accuracy, efficiency, and coordination
• TeleMoM integrates multiple LLMs for better Telecom decisions

Innovation

• Consensus-driven ensemble framework for Telecom
• Two-stage process with proponent and adjudicator models
• Quality-checking mechanism enhances accuracy and reduces biases