MoD: A Distribution-Based Approach for Merging Large Language Models

Quy-Anh Dang, Chris Ngo

Keywords: LLMs Merging

TL;DR: Merging LLMs that operates directly on their output probability distributions, rather than on model weight

Abstract:

Large language models (LLMs) have enabled the development of numerous specialized, task-specific variants. However, the maintenance and deployment of these individual models present substantial challenges in terms of resource utilization and operational efficiency. In this work, we propose the \textit{Mixture of Distributions (MoD)} framework, a novel approach for merging LLMs that operates directly on their output probability distributions, rather than on model weights. Unlike traditional weight-averaging methods, MoD effectively preserves the specialized capabilities of individual models while enabling efficient knowledge sharing across tasks. Through extensive experimentation on mathematical reasoning benchmarks using Qwen2.5 models, we demonstrate that MoD significantly outperforms existing model merging techniques across multiple benchmarks. All code, data, and experimental materials are published at \url{https://github.com/knovel-eng/mod}.

Submission Number: 1