Bayesian Network Fusion of Large Language Models for Sentiment Analysis

2025-10-30

AI Summary

To address the limited interpretability, unstable cross-domain performance, and high computational cost of large language models (LLMs) in sentiment analysis, this paper proposes the Bayesian Network-based LLM Fusion framework (BNLF). BNLF models the probabilistic outputs of FinBERT, RoBERTa, and BERTweet as nodes in a Bayesian network and performs interpretable late fusion via probabilistic inference, thereby enhancing robustness to distributional shifts. Experiments on three annotated financial-domain corpora demonstrate that BNLF achieves an average accuracy improvement of approximately 6% over baseline methods, while providing explicit, traceable probabilistic explanations for predictions. To our knowledge, this is the first work to integrate structured probabilistic graphical models into multi-LLM sentiment ensembles—effectively balancing performance gains with decision transparency. BNLF establishes a novel paradigm for resource-efficient, high-fidelity cross-domain sentiment analysis.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large language models (LLMs) continue to advance, with an increasing number of domain-specific variants tailored for specialised tasks. However, these models often lack transparency and explainability, can be costly to fine-tune, require substantial prompt engineering, yield inconsistent results across domains, and impose significant adverse environmental impact due to their high computational demands. To address these challenges, we propose the Bayesian network LLM fusion (BNLF) framework, which integrates predictions from three LLMs, including FinBERT, RoBERTa, and BERTweet, through a probabilistic mechanism for sentiment analysis. BNLF performs late fusion by modelling the sentiment predictions from multiple LLMs as probabilistic nodes within a Bayesian network. Evaluated across three human-annotated financial corpora with distinct linguistic and contextual characteristics, BNLF demonstrates consistent gains of about six percent in accuracy over the baseline LLMs, underscoring its robustness to dataset variability and the effectiveness of probabilistic fusion for interpretable sentiment classification.

Problem

• Enhancing LLM transparency and explainability for sentiment analysis
• Reducing computational costs and environmental impact of LLMs
• Improving cross-domain consistency in sentiment classification results

Innovation

• Bayesian network fuses multiple LLM predictions
• Probabilistic late fusion enhances sentiment analysis accuracy
• Framework integrates FinBERT RoBERTa BERTweet via nodes