Example DocumentsLearning to Generate Reviews and Discovering SentimentThis paper explores the properties of byte-level recurrent language models and finds a single unit that performs sentiment analysis. It demonstrates that the sentiment un...ListenView DetailsNeural Turing MachinesThe paper introduces Neural Turing Machines (NTMs), which augment neural networks with external memory resources that they can interact with through attentional processes...ListenView DetailsDeep Learning is Not So Mysterious or DifferentThis paper argues that anomalous generalization behaviors like benign overfitting and double descent are not unique to deep neural networks and can be understood through ...ListenView DetailsScaling Laws for Neural Language ModelsThis paper studies empirical scaling laws for language model performance on the cross-entropy loss, showing that the loss scales as a power-law with model size, dataset s...ListenView DetailsVARIATIONAL LOSSY AUTOENCODERThe paper introduces Variational Lossy Autoencoder(VLAE), a simple and principled method to learn such global representations by combining Variational Autoencoder (VAE) w...ListenView DetailsListen to example documents converted to MP3