1 paper
Miao Lu, Zeyu Bian, Kaiyue Wen +4
The empirical success of pretraining large language models has inspired a deeper investigation into the underlying loss landscapes and the optimization dynamics. Recent empirical a…