Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code.

/pdf/roberta-a-robustly-optimized-bert-pretraining-approach-2rj0bdyhim.pdf

RoBERTa: A Robustly Optimized BERT Pretraining Approach

Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Machine learning

While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train.

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

Proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function. Through an enormous experimental effort1–4, the structures of around 100,000 unique proteins have been determined5, but this represents a small fraction of the billions of known protein sequences6,7. Structural coverage is bottlenecked by the months to years of painstaking effort required to determine a single protein structure. Accurate computational approaches are needed to address this gap and to enable large-scale structural bioinformatics. Predicting the three-dimensional structure that a protein will adopt based solely on its amino acid sequence—the structure prediction component of the ‘protein folding problem’8—has been an important open research problem for more than 50 years9. Despite recent progress10–14, existing methods fall far short of atomic accuracy, especially when no homologous structure is available. Here we provide the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known. We validated an entirely redesigned version of our neural network-based model, AlphaFold, in the challenging 14th Critical Assessment of protein Structure Prediction (CASP14)15, demonstrating accuracy competitive with experimental structures in a majority of cases and greatly outperforming other methods. Underpinning the latest version of AlphaFold is a novel machine learning approach that incorporates physical and biological knowledge about protein structure, leveraging multi-sequence alignments, into the design of the deep learning algorithm. AlphaFold predicts protein structures with an accuracy competitive with experimental structures in the majority of cases using a novel deep learning architecture.

/pdf/highly-accurate-protein-structure-prediction-with-alphafold-2xjbzfwu3n.pdf

Highly accurate protein structure prediction with AlphaFold

Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.

/pdf/language-models-are-few-shot-learners-2fg8gvia7m.pdf

Language Models are Few-Shot Learners

The LinGO Redwoods initiative is a seed activity in the design and development of a new type of treebank. While several medium- to large-scale treebanks exist for English (and for other major languages), pre-existing publicly available resources exhibit the following limitations: (i) annotation is mono-stratal, either encoding topological (phrase structure) or tectogrammatical (dependency) information, (ii) the depth of linguistic information recorded is comparatively shallow, (iii) the design and format of linguistic representation in the treebank hard-wires a small, predefined range of ways in which information can be extracted from the treebank, and (iv) representations in existing treebanks are static and over the (often year- or decade-long) evolution of a large-scale treebank tend to fall behind the development of the field. LinGO Redwoods aims at the development of a novel treebanking methodology, rich in nature and dynamic both in the ways linguistic data can be retrieved from the treebank in varying granularity and in the constant evolution and regular updating of the treebank itself. Since October 2001, the project is working to build the foundations for this new type of treebank, to develop a basic set of tools for treebank construction and maintenance, and to construct an initial set of 10,000 annotated trees to be distributed together with the tools under an open-source license.

/pdf/the-lingo-redwoods-treebank-motivation-and-preliminary-3s5gpmknyb.pdf

The LinGO Redwoods treebank motivation and preliminary applications

We improve the quality of statistical machine translation (SMT) by applying models that predict word forms from their stems using extensive morphological and syntactic information from both the source and target languages. Our inflection generation models are trained independently of the SMT system. We investigate different ways of combining the inflection prediction component with the SMT system by training the base MT system on fully inflected forms or on word stems. We applied our inflection generation models in translating English into two morphologically complex languages, Russian and Arabic, and show that our model improves the quality of SMT over both phrasal and syntax-based SMT systems according to BLEU and human judgements.

/pdf/applying-morphology-generation-models-to-machine-translation-3c8wb5i5t0.pdf

Applying Morphology Generation Models to Machine Translation

We present a novel Bayesian model for semi-supervised part-of-speech tagging. Our model extends the Latent Dirichlet Allocation model and incorporates the intuition that words' distributions over tags, p(t|w), are sparse. In addition we introduce a model for determining the set of possible tags of a word which captures important dependencies in the ambiguity classes of words. Our model outperforms the best previously proposed model for this task on a standard dataset.

/pdf/a-bayesian-lda-based-model-for-semi-supervised-part-of-2fjh6ggu2a.pdf

A Bayesian LDA-based model for semi-supervised part-of-speech tagging

Representing documents by vectors that are independent of language enhances machine translation and multilingual text categorization. We use discriminative training to create a projection of documents from multiple languages into a single translingual vector space. We explore two variants to create these projections: Oriented Principal Component Analysis (OPCA) and Coupled Probabilistic Latent Semantic Analysis (CPLSA). Both of these variants start with a basic model of documents (PCA and PLSA). Each model is then made discriminative by encouraging comparable document pairs to have similar vector representations. We evaluate these algorithms on two tasks: parallel document retrieval for Wikipedia and Europarl documents, and cross-lingual text classification on Reuters. The two discriminative variants, OPCA and CPLSA, significantly outperform their corresponding baselines. The largest differences in performance are observed on the task of retrieval when the documents are only comparable and not parallel. The OPCA method is shown to perform best.

/pdf/translingual-document-representations-from-discriminative-37dpv6gce4.pdf

Translingual Document Representations from Discriminative Projections

We present a novel method for predicting inflected word forms for generating morphologically rich languages in machine translation We utilize a rich set of syntactic and morphological knowledge sources from both source and target sentences in a probabilistic model, and evaluate their contribution in generating Russian and Arabic sentences Our results show that the proposed model substantially outperforms the commonly used baseline of a trigram target language model; in particular, the use of morphological and syntactic features leads to large gains in prediction accuracy We also show that the proposed method is effective with a relatively small amount of data

/pdf/generating-complex-morphology-for-machine-translation-1x2c86f7ds.pdf

Kristina Toutanova

Papers

The LinGO Redwoods treebank motivation and preliminary applications

Applying Morphology Generation Models to Machine Translation

A Bayesian LDA-based model for semi-supervised part-of-speech tagging

Translingual Document Representations from Discriminative Projections

Generating Complex Morphology for Machine Translation