Model Merging involves combining the weights of two or more fine-tuned models to create a single model that retains the skills of both, without retraining.
Exploded in the open-source community.
Methods like SLERP and DARE are standards on Hugging Face.