Merging large language models (LLMs) is a fascinating and surprisingly efficient way to create high-performing LLM models without the...