Large language models (LLMs) excel at few-shot in-context learning (ICL) -- learning from a few examples provided in context at inference, without any weight updates. Newly expanded context windows allow us to investigate ICL with hundreds or thousands of examples – the many-shot regime. Going from few-shot to many-shot, we observe significant performance gains across a wide variety of generative and discriminative tasks. While promising, many-shot ICL can be bottlenecked by the available amount of human-generated outputs. To mitigate this limitation, we explore two new settings: (1) "Reinforced ICL" that uses model-generated chain-of-thought rationales in place of human rationales, and (2) "Unsupervised ICL" where we remove rationales from the prompt altogether, and prompts the model only with domain-specific inputs. We find that both Reinforced and Unsupervised ICL can be quite effective in the many-shot regime, particularly on complex reasoning tasks. We demonstrate that, unlike few-shot learning, many-shot learning is effective at overriding pretraining biases, can learn high-dimensional functions with numerical inputs, and performs comparably to supervised fine-tuning. Finally, we reveal the limitations of next-token prediction loss as an indicator of downstream ICL performance.

Abstractive summarization has been studied using neural sequence transduction methods with datasets of large, paired document-summary examples. However, such datasets are rare and the models trained from them do not generalize to other domains. Recently, some progress has been made in learning sequence-to-sequence mappings with only unpaired examples. In our work, we consider the setting where there are only documents (product or business reviews) with no summaries provided, and propose an end-to-end, neural model architecture to perform unsupervised abstractive summarization. Our proposed model consists of an auto-encoder where the mean of the representations of the input reviews decodes to a reasonable summary-review. We consider variants of the proposed architecture and perform an ablation study to show the importance of specific components. We show through metrics and human evaluation that the generated summaries are highly abstractive, fluent, relevant, and representative of the average sentiment of the input reviews. Finally, we collect a ground-truth evaluation dataset and show that our model outperforms a strong extractive baseline.

The ability to infer persona from dialogue can have applications in areas ranging from computational narrative analysis to personalized dialogue generation.We introduce neural models to learn persona embeddings in a supervised character trope classification task.The models encode dialogue snippets from IMDB into representations that can capture the various categories of film characters.The bestperforming models use a multi-level attention mechanism over a set of utterances.We also utilize prior knowledge in the form of textual descriptions of the different tropes.We apply the learned embeddings to find similar characters across different movies, and cluster movies according to the distribution of the embeddings.The use of short conversational text as input, and the ability to learn from prior knowledge using memory, suggests these methods could be applied to other domains.

Stories can have tremendous power - not only useful for entertainment, they can activate our interests and mobilize our actions. The degree to which a story resonates with its audience may be in part reflected in the emotional journey it takes the audience upon. In this paper, we use machine learning methods to construct emotional arcs in movies, calculate families of arcs, and demonstrate the ability for certain arcs to predict audience engagement. The system is applied to Hollywood films and high quality shorts found on the web. We begin by using deep convolutional neural networks for audio and visual sentiment analysis. These models are trained on both new and existing large-scale datasets, after which they can be used to compute separate audio and visual emotional arcs. We then crowdsource annotations for 30-second video clips extracted from highs and lows in the arcs in order to assess the micro-level precision of the system, with precision measured in terms of agreement in polarity between the system's predictions and annotators' ratings. These annotations are also used to combine the audio and visual predictions. Next, we look at macro-level characterizations of movies by investigating whether there exist 'universal shapes' of emotional arcs. In particular, we develop a clustering approach to discover distinct classes of emotional arcs. Finally, we show on a sample corpus of short web videos that certain emotional arcs are statistically significant predictors of the number of comments a video receives. These results suggest that the emotional arcs learned by our approach successfully represent macroscopic aspects of a video story that drive audience engagement. Such machine understanding could be used to predict audience reactions to video stories, ultimately improving our ability as storytellers to communicate with each other.

A common criticism of database systems is that they are hard to query for users uncomfortable with a formal query language. To address this problem, form-based interfaces and keyword search have been proposed; while both have benefits, both also have limitations. In this paper, we investigate combining the two with the hopes of creating an approach that provides the best of both. Specifically, we propose to take as input a target database and then generate and index a set of query forms offline. At query time, a user with a question to be answered issues standard keyword search queries; but instead of returning tuples, the system returns forms relevant to the question. The user may then build a structured query with one of these forms and submit it back to the system for evaluation. In this paper, we address challenges that arise in form generation, keyword search over forms, and ranking and displaying these forms. We explore techniques to tackle these challenges, and present experimental results suggesting that the approach of combining keyword search and form-based interfaces is promising.

A “sparse ” data set typically has hundreds or even thousands of attributes, but most objects have non-null values for only a small number of these attributes. A popular view about sparse data is that it arises merely as the result of poor schema design. In this paper, we argue that rather than being the result of inept schema design, storing a sparse data set in a single table is the right way to proceed. However, for this to be the case, RDBMSs must provide sparse data management facilities that go beyond the previously studied requirement of storing such data sets efficiently. In particular, an RDBMS must 1) enable users to effectively build ad hoc queries over a very large number of attributes, and 2) support efficient evaluation of these queries over a wide, sparse table. We propose techniques that provide these capabilities, and argue that the single-table approach is a necessary component of selfmanaging database systems because it frees users from a tedious and potentially ineffective schema-design phase when managing sparse data sets.

The area of automatic selection of physical database design to optimize the performance of a relational database system based on a workload of SQL queries and updates has gained prominence in recent years. Major database vendors have released automated physical database design tools with the goal of reducing the total cost of ownership. An important assumption underlying these tools is that the workload is a set of SQL statements. In this paper, we show that being able to treat the workload as a sequence, i.e., exploiting the ordering of statements can significantly broaden the usage of such tools. We present scenarios where exploiting sequence information in the workload is crucial for performance tuning. We also propose techniques for addressing the technical challenges arising from treating the workload as a sequence. We evaluate the effectiveness of our techniques through experiments on Microsoft SQL Server.