#ICLR #TrainBetterLM I am at ICLR, come to our posters for improved language model training! Recycle gradients for faster neural net training with AdEMAmix iclr.cc/virtual/2025... (Fri Apr 25, 10 am). 1/3