@article{trainingoverheadratioapractical, title = {Training Overhead Ratio: A Practical Reliability Metric for Large Language Model Training Systems}, author = {Ning Lu and Qian Xie and Hao Zhang and Wenyi Fang and Yang Zheng and Zheng Hu and Jiantao Ma}, year = {2024}, eprint = {2408.07482}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2408.07482v3}, }