cs.LGAug 29, 2025

Rethinking Layer-wise Model Merging through Chain of Merges

Authors: Pietro Buzzega, Riccardo Salami, Angelo Porrello, Simone Calderara

Organizations: AImageLab, Department of Engineering, University of Modena and Reggio Emilia, Italy.

Abstract

Model merging has emerged as a simple yet effective method to fuse models without re-training. Existing techniques operate at the level of individual layers, thereby overlooking the inter-layer dependencies inherent in deep networks. We show that this simplification induces distributional mismatches in intermediate activations during merging, as changes applied to early layers fail to propagate to downstream ones. We identify these mismatches as a form of covariate shift that compounds across the network. To address this issue, we propose Chain of Merges (CoM), a novel procedure that sequentially merges weights across layers while accounting for inter-layer interactions. In particular, CoM mitigates covariate shift through a series of regression problems, where input activations are recomputed at each step to reflect the updated representations. Additionally, we introduce a dynamic weighting mechanism that prioritizes task-layer pairs most susceptible to performance degradation, improving robustness during merging. Experiments on standard benchmarks demonstrate that CoM achieves state-of-the-art performance across heterogeneous tasks.

Figures & tables

Appendix figures & tables13 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

May 2, 2026

Model Merging: Foundations and Algorithms

Modern deep learning usually treats models as separate artifacts: trained independently, specialized for particular purposes, and replaced when improved versions appear. This thesis studies model merging as an alternative paradigm: combining independently trained neural networks directly in weight space, with little or no optimization and without requiring access to the original training data. The thesis considers two main regimes. In the single-task setting, where models share an objective but differ in initialization, we introduce C2^2M3^3, a cycle-consistent merging algorithm based on Frank-Wolfe optimization. C2^2M3^3 aligns multiple networks into a shared, reference-free parameter space, making weight averaging meaningful without privileging any individual model. In the multi-task setting, where models are fine-tuned for different downstream tasks from a common pretrained initialization, we first develop a theoretical account of task vectors as approximate gradients. This explains both the effectiveness and the limitations of task arithmetic. Building on this view, we show that task vectors inherit the low-rank structure of gradients and introduce Task Singular Vectors (TSV), a decomposition that enables compression and interference reduction through TSV-Merge. We then present MASS, an input-adaptive routing method that uses TSV geometry to select task-relevant subspaces at inference time. Finally, we introduce MERGE3^3, an evolutionary merging framework that uses Item Response Theory to reduce evaluation costs by up to 50×\times while preserving solution quality. Together, these contributions provide theoretical and algorithmic foundations for model merging, supporting a paradigm in which learned capabilities can be composed, reused, and extended across models.
Oct 6, 2026

A Broader Look at Model Merging: Rethinking Implicit Regularization Induced by Task Arithmetic

Model merging aims to build a multi-task model cheaply by combining the weights of individual task-specific models. To perform well across multiple tasks, most existing merging methods use an additional dataset to find the coefficients for the best linear combination of task-specific weight updates. However, we identify an implicit regularization in this standard practice: searching over coefficients restricts the candidate models to a subspace spanned by task-specific weight updates. In this work, we investigate whether this regularization is actually useful. Surprisingly, empirical results show that optimizing merged-model weights without this regularization significantly boosts the performance of common merging methods across multiple architectures, domains, and even in an extremely data-limited scenario where only one instance is available per class. Moreover, directly optimizing the pretrained model weights even outperforms some existing merging methods. Analysis shows that better multi-task weights exist outside the subspace and can be found using multiple methods. We study different strategies for using the additional dataset, discussing their practical use and implications for model merging. Overall, this work calls for revisiting the existing model-merging pipeline, motivating a broader exploration of the weight space and a reconsideration of the implicit regularization induced by task arithmetic.
Apr 1, 2026

Model Merging via Data-Free Covariance Estimation

Model merging provides a way of cheaply combining individual models to produce a model that inherits each individual's capabilities. While some merging methods can approach the performance of multitask training, they are often heuristically motivated and lack theoretical justification. A principled alternative is to pose model merging as a layer-wise optimization problem that directly minimizes interference between tasks. However, this formulation requires estimating per-layer covariance matrices from data, which may not be available when performing merging. In contrast, many of the heuristically-motivated methods do not require auxiliary data, making them practically advantageous. In this work, we revisit the interference minimization framework and show that, under certain conditions, covariance matrices can be estimated directly from difference matrices, eliminating the need for data while also reducing computational costs. We validate our approach across vision and language benchmarks on models ranging from 86M parameters to 7B parameters, outperforming previous data-free state-of-the-art merging methods