InfiFPO: Implicit Model Fusion via Preference Optimization in Large Language Models

TL;DR

InfiFPO introduces an implicit model fusion method that leverages preference optimization to enhance large language models by synthesizing multi-source probabilities, outperforming existing fusion techniques across various benchmarks.

Contribution

The paper proposes InfiFPO, a novel preference optimization approach that fuses multiple models implicitly by synthesizing probabilities, addressing limitations of previous methods that only used response outputs.

Findings

InfiFPO outperforms existing methods on 11 benchmarks.

Using Phi-4 as pivot, InfiFPO improves performance from 79.95 to 83.33.

InfiFPO enhances capabilities in mathematics, coding, and reasoning tasks.

Abstract

Model fusion combines multiple Large Language Models (LLMs) with different strengths into a more powerful, integrated model through lightweight training methods. Existing works on model fusion focus primarily on supervised fine-tuning (SFT), leaving preference alignment (PA) --a critical phase for enhancing LLM performance--largely unexplored. The current few fusion methods on PA phase, like WRPO, simplify the process by utilizing only response outputs from source models while discarding their probability information. To address this limitation, we propose InfiFPO, a preference optimization method for implicit model fusion. InfiFPO replaces the reference model in Direct Preference Optimization (DPO) with a fused source model that synthesizes multi-source probabilities at the sequence level, circumventing complex vocabulary alignment challenges in previous works and meanwhile maintaining the probability information. By introducing probability clipping and max-margin fusion strategies, InfiFPO enables the pivot model to align with human preferences while effectively distilling knowledge from source models. Comprehensive experiments on 11 widely-used benchmarks demonstrate that InfiFPO consistently outperforms existing model fusion and preference optimization methods. When using Phi-4 as the pivot model, InfiFPO improve its average performance from 79.95 to 83.33 on 11 benchmarks, significantly improving its capabilities in mathematics, coding, and reasoning tasks.