Saved in:
Bibliographic Details
Main Authors: Wu, Chuhan, Tang, Ruiming
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2408.09895
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914948573036544
author Wu, Chuhan
Tang, Ruiming
author_facet Wu, Chuhan
Tang, Ruiming
contents 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 of loss, which is influenced by various factors such as model architectures, data distributions, tokenizers, and computation precision. Thus, estimating the real performance of LLMs with different training settings rather than loss may be quite useful in practical development. In this article, we present an empirical equation named "Performance Law" to directly predict the MMLU score of an LLM, which is a widely used metric to indicate the general capability of LLMs in real-world conversations and applications. Based on only a few key hyperparameters of the LLM architecture and the size of training data, we obtain a quite accurate MMLU prediction of various LLMs with diverse sizes and architectures developed by different organizations in different years. Performance law can be used to guide the choice of LLM architecture and the effective allocation of computational resources without extensive experiments.
format Preprint
id arxiv_https___arxiv_org_abs_2408_09895
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Performance Law of Large Language Models
Wu, Chuhan
Tang, Ruiming
Computation and Language
Machine Learning
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 of loss, which is influenced by various factors such as model architectures, data distributions, tokenizers, and computation precision. Thus, estimating the real performance of LLMs with different training settings rather than loss may be quite useful in practical development. In this article, we present an empirical equation named "Performance Law" to directly predict the MMLU score of an LLM, which is a widely used metric to indicate the general capability of LLMs in real-world conversations and applications. Based on only a few key hyperparameters of the LLM architecture and the size of training data, we obtain a quite accurate MMLU prediction of various LLMs with diverse sizes and architectures developed by different organizations in different years. Performance law can be used to guide the choice of LLM architecture and the effective allocation of computational resources without extensive experiments.
title Performance Law of Large Language Models
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2408.09895