The mixture proportions of pretraining data domains (e.g., Wikipedia, bo...
Reward design in reinforcement learning (RL) is challenging since specif...
Selecting a suitable training dataset is crucial for both general-domain...
Language models (LMs) are becoming the foundation for almost all major
l...
Language modeling on large-scale datasets leads to impressive performanc...
We consider unsupervised domain adaptation (UDA), where labeled data fro...
Machine learning systems deployed in the wild are often trained on a sou...
Large pretrained language models such as GPT-3 have the surprising abili...
Out-of-distribution detection is an important component of reliable ML
s...
Pretrained language models have achieved state-of-the-art performance wh...
Distribution shifts can cause significant degradation in a broad range o...
Consider a prediction setting where a few inputs (e.g., satellite images...
We focus on prediction problems with high-dimensional outputs that are
s...
Adversarial training augments the training set with perturbations to imp...
While adversarial training can improve robust accuracy (against an
adver...
Many machine learning tasks require sampling a subset of items from a
co...
Large amounts of labeled data are typically required to train deep learn...