“Model Merging by Uncertainty-Based Gradient Matching”, Nico Daheim, Thomas Möllenhoff, Edoardo Maria Ponti, Iryna Gurevych, Mohammad Emtiyaz Khan2023-10-19 (Transformer, Bayes)⁠:

Models trained on different datasets can be merged by a weighted-averaging of their parameters, but why does it work and when can it fail?

Here, we connect the inaccuracy of weighted-averaging to mismatches in the gradients and propose a new uncertainty-based scheme to improve the performance by reducing the mismatch. The connection also reveals implicit assumptions in other schemes such as averaging, task arithmetic, and Fisher-weighted averaging.

Our new method gives consistent improvements for large language models and vision transformers, both in terms of performance and robustness to hyperparameters.