@article{datasetdecompositionfasterllmtraining, title = {Dataset Decomposition: Faster LLM Training with Variable Sequence Length Curriculum}, author = {Hadi Pouransari and Chun-Liang Li and Jen-Hao Rick Chang and Pavan Kumar Anasosalu Vasu and Cem Koc and Vaishaal Shankar and Oncel Tuzel}, year = {2024}, eprint = {2405.13226}, archivePrefix = {arXiv}, url = {https://arxiv.org/abs/2405.13226v2}, }