Reinforcement learning from human feedback (RLHF) is a technique for tra...
We propose the Thinker algorithm, a novel approach that enables reinforc...
Visual object recognition systems need to generalize from a set of 2D
tr...
Manipulation is a common concern in many domains, such as social media,
...
A principled understanding of generalization in deep learning may requir...
Research in Fairness, Accountability, Transparency, and Ethics (FATE) ha...
Current state-of-the-art deep networks are all powered by backpropagatio...
Reward functions are notoriously difficult to specify, especially for ta...
Existing offline reinforcement learning (RL) algorithms typically assume...
Neural networks are known to be biased towards learning mechanisms that ...
We present a smoothly broken power law functional form that accurately m...
Adversarial robustness continues to be a major challenge for deep learni...
We provide the first formal definition of reward hacking, a phenomenon w...
Modern machine learning research relies on relatively few carefully cura...
Learning models that generalize under different distribution shifts in
m...
The range of application of artificial intelligence (AI) is vast, as is ...
Active reinforcement learning (ARL) is a variant on reinforcement learni...
Decisions made by machine learning systems have increasing influence on ...
Framed in positive terms, this report examines how technical AI research...
With the recent wave of progress in artificial intelligence (AI) has com...
Generalizing outside of the training distribution is an open challenge f...
One obstacle to applying reinforcement learning algorithms to real-world...
Using variational Bayes neural networks, we develop an algorithm capable...
Normalizing flows and autoregressive models have been successfully combi...
We propose Nested LSTMs (NLSTM), a novel RNN architecture with multiple
...
The recent literature on deep learning offers new tools to learn a rich
...
We propose Bayesian hypernetworks: a framework for approximate Bayesian
...
We examine the role of memorization in deep learning, drawing connection...
We propose zoneout, a novel method for regularizing RNNs. At each timest...
We stabilize the activations of Recurrent Neural Networks (RNNs) by
pena...
Regularized training of an autoencoder typically results in hidden unit
...