The impact of random seeds in fine-tuning large language models (LLMs) has been largely overlooked despite its potential influence on model performance.In this study, we systematically evaluate the effects of random seeds on LLMs using the GLUE and SuperGLUE benchmarks. We analyze the macro-level impact through traditional metrics like accuracy and F1, calculating their mean and variance to quantify performance fluctuations. To capture the micro-level effects, we introduce a novel metric, consistency, measuring the stability of individual predictions across runs. Our experiments reveal significant variance at both macro and micro levels, underscoring the need for careful consideration of random seeds in fine-tuning and evaluation.
@article{arxiv.2503.07329,
title = {Assessing the Macro and Micro Effects of Random Seeds on Fine-Tuning Large Language Models},
author = {Nghia Bui and Guergana Savova and Lijing Wang},
journal= {arXiv preprint arXiv:2503.07329},
year = {2025}
}
Comments
7 pages, 5 tables, 3 figures. Accepted at IJCNLP 2025. This is the final, peer-reviewed version of the work, which supersedes and extends the unauthorized draft previously posted as arXiv:2503.07329