Model Merging via Data-Free Covariance Estimation
This work addresses the practical challenge of merging models when data is unavailable, offering a more efficient and data-free solution for combining individual models' capabilities.
The paper tackles the problem of model merging without needing auxiliary data by estimating covariance matrices from difference matrices, eliminating data requirements and reducing computational costs. The approach outperforms previous data-free state-of-the-art methods on vision and language benchmarks with models ranging from 86M to 7B parameters.
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