Choose one model and fine-tune it to create a gap between the original model and the fine-tuned one. It doesn't matter whether the evaluation score is higher or lower.
Merge the two models.
Evaluate the merged model.
Fine-tune a specific evaluation part of the model if you need to increase the score for that part. (It's unlikely to work as you think, but you can try it.)
Merge the models again.
Evaluate again.
Keep going until the average evaluation score is higher than the original one.
That's it. Simple.
You can create a framework to automate this process.
