Resolving Interference (RI): Disentangling Models for Improved Model Merging

TL;DR

This paper introduces Resolving Interference (RI), a lightweight method that reduces cross-task interference in model merging by disentangling models using unlabeled data, leading to improved performance and robustness.

Contribution

The paper proposes RI, a novel disentanglement framework that minimizes interference during model merging without requiring task-specific data, enhancing performance and generalization.

Findings

RI improves merging performance by up to 3.8%.

RI enhances generalization to unseen domains by up to 2.3%.

RI is robust to auxiliary data sources and less sensitive to hyperparameters.

Abstract

Model merging has shown that multitask models can be created by directly combining the parameters of different models that are each specialized on tasks of interest. However, models trained independently on distinct tasks often exhibit interference that degrades the merged model's performance. To solve this problem, we formally define the notion of Cross-Task Interference as the drift in the representation of the merged model relative to its constituent models. Reducing cross-task interference is key to improving merging performance. To address this issue, we propose our method, Resolving Interference (RI), a light-weight adaptation framework which disentangles expert models to be functionally orthogonal to the space of other tasks, thereby reducing cross-task interference. RI does this whilst using only unlabeled auxiliary data as input (i.e., no task-data is needed), allowing it to be applied in data-scarce scenarios. RI consistently improves the performance of state-of-the-art merging methods by up to 3.8% and generalization to unseen domains by up to 2.3%. We also find RI to be robust to the source of auxiliary input while being significantly less sensitive to tuning of merging hyperparameters. Our codebase is available at: https://github.com/pramesh39/resolving_interference