Performance Law of Large Language Models

Chuhan Wu, Ruiming Tang

Guided by the belief of the scaling law, large language models (LLMs) have achieved impressive performance in recent years. However, scaling law only gives a qualitative estimation…