Commonsense reasoning is intuitive for humans but has been a long-term challenge for artificial intelligence (AI). Recent advancements in pretrained language models have shown promising results on several commonsense benchmark datasets. However, the reliability and comprehensiveness of these benchmarks towards assessing model's commonsense reasoning ability remains unclear. To this end, we introduce a new commonsense reasoning benchmark dataset comprising natural language true/false statements, with each sample paired with its complementary counterpart, resulting in 4k sentence pairs. We propose a pairwise accuracy metric to reliably measure an agent's ability to perform commonsense reasoning over a given situation. The dataset is crowdsourced and enhanced with an adversarial model-in-the-loop setup to incentivize challenging samples. To facilitate a systematic analysis of commonsense capabilities, we design our dataset along the dimensions of knowledge domains, reasoning scenarios and numeracy. Experimental results demonstrate that our strongest baseline (UnifiedQA-3B), after fine-tuning, achieves ~71% standard accuracy and ~51% pairwise accuracy, well below human performance (~95% for both metrics). The dataset is available at https://github.com/PlusLabNLP/Com2Sense.

The goal of controlled feature selection is to discover the features a response depends on while limiting the proportion of false discoveries to a predefined level. Recently, multiple methods have been proposed that use deep learning to generate knockoffs for controlled feature selection through the Model-X knockoff framework. We demonstrate, however, that these methods often fail to control the false discovery rate (FDR). There are two reasons for this shortcoming. First, these methods often learn inaccurate models of features. Second, the "swap" property, which is required for knockoffs to be valid, is often not well enforced. We propose a new procedure called FlowSelect that remedies both of these problems. To more accurately model the features, FlowSelect uses normalizing flows, the state-of-the-art method for density estimation. To circumvent the need to enforce the swap property, FlowSelect uses a novel MCMC-based procedure to directly compute p-values for each feature. Asymptotically, FlowSelect controls the FDR exactly. Empirically, FlowSelect controls the FDR well on both synthetic and semi-synthetic benchmarks, whereas competing knockoff-based approaches fail to do so. FlowSelect also demonstrates greater power on these benchmarks. Additionally, using data from a genome-wide association study of soybeans, FlowSelect correctly infers the genetic variants associated with specific soybean traits.

Over the past decades, numerous practical applications of machine learning techniques have shown the potential of data-driven approaches in a large number of computing fields. Machine learning is increasingly included in computing curricula in higher education, and a quickly growing number of initiatives are expanding it in K-12 computing education, too. As machine learning enters K-12 computing education, understanding how intuition and agency in the context of such systems is developed becomes a key research area. But as schools and teachers are already struggling with integrating traditional computational thinking and traditional artificial intelligence into school curricula, understanding the challenges behind teaching machine learning in K-12 is an even more daunting challenge for computing education research. Despite the central position of machine learning in the field of modern computing, the computing education research body of literature contains remarkably few studies of how people learn to train, test, improve, and deploy machine learning systems. This is especially true of the K-12 curriculum space. This article charts the emerging trajectories in educational practice, theory, and technology related to teaching machine learning in K-12 education. The article situates the existing work in the context of computing education in general, and describes some differences that K-12 computing educators should take into account when facing this challenge. The article focuses on key aspects of the paradigm shift that will be required in order to successfully integrate machine learning into the broader K-12 computing curricula. A crucial step is abandoning the belief that rule-based "traditional" programming is a central aspect and building block in developing next generation computational thinking.

Human activity recognition in videos is a challenging problem that has drawn a lot of interest, particularly when the goal requires the analysis of a large video database. AOLME project provides a collaborative learning environment for middle school students to explore mathematics, computer science, and engineering by processing digital images and videos. As part of this project, around 2200 hours of video data was collected for analysis. Because of the size of the dataset, it is hard to analyze all the videos of the dataset manually. Thus, there is a huge need for reliable computer-based methods that can detect activities of interest. My thesis is focused on the development of accurate methods for detecting and tracking objects in long videos. All the models are validated on videos from 7 different sessions, ranging from 45 minutes to 90 minutes. The keyboard detector achieved a very high average precision (AP) of 92% at 0.5 intersection over union (IoU). Furthermore, a combined system of the detector with a fast tracker KCF (159fps) was developed so that the algorithm runs significantly faster without sacrificing accuracy. For a video of 23 minutes having resolution 858X480 @ 30 fps, the detection alone runs at 4.7Xthe real-time, and the combined algorithm runs at 21Xthe real-time for an average IoU of 0.84 and 0.82, respectively. The hand detector achieved average precision (AP) of 72% at 0.5 IoU. The detection results were improved to 81% using optimal data augmentation parameters. The hand detector runs at 4.7Xthe real-time with AP of 81% at 0.5 IoU. The hand detection method was integrated with projections and clustering for accurate proposal generation. This approach reduced the number of false-positive hand detections by 80%. The overall hand detection system runs at 4Xthe real-time, capturing all the activity regions of the current collaborative group.

Classifiers commonly make use of pre-annotated datasets, wherein a model is evaluated by pre-defined metrics on a held-out test set typically made of human-annotated labels. Metrics used in these evaluations are tied to the availability of well-defined ground truth labels, and these metrics typically do not allow for inexact matches. These noisy ground truth labels and strict evaluation metrics may compromise the validity and realism of evaluation results. In the present work, we discuss these concerns and conduct a systematic posthoc verification experiment on the entity linking (EL) task. Unlike traditional methodologies, which asks annotators to provide free-form annotations, we ask annotators to verify the correctness of annotations after the fact (i.e., posthoc). Compared to pre-annotation evaluation, state-of-the-art EL models performed extremely well according to the posthoc evaluation methodology. Posthoc validation also permits the validation of the ground truth dataset. Surprisingly, we find predictions from EL models had a similar or higher verification rate than the ground truth. We conclude with a discussion on these findings and recommendations for future evaluations.

Language modeling (LM) for automatic speech recognition (ASR) does not usually incorporate utterance level contextual information. For some domains like voice assistants, however, additional context, such as the time at which an utterance was spoken, provides a rich input signal. We introduce an attention mechanism for training neural speech recognition language models on both text and non-linguistic contextual data. When applied to a large de-identified dataset of utterances collected by a popular voice assistant platform, our method reduces perplexity by 7.0% relative over a standard LM that does not incorporate contextual information. When evaluated on utterances extracted from the long tail of the dataset, our method improves perplexity by 9.0% relative over a standard LM and by over 2.8% relative when compared to a state-of-the-art model for contextual LM.

Learning to reach goal states and learning diverse skills through mutual information (MI) maximization have been proposed as principled frameworks for self-supervised reinforcement learning, allowing agents to acquire broadly applicable multitask policies with minimal reward engineering. Starting from a simple observation that the standard goal-conditioned RL (GCRL) is encapsulated by the optimization objective of variational empowerment, we discuss how GCRL and MI-based RL can be generalized into a single family of methods, which we name variational GCRL (VGCRL), interpreting variational MI maximization, or variational empowerment, as representation learning methods that acquire functionally-aware state representations for goal reaching. This novel perspective allows us to: (1) derive simple but unexplored variants of GCRL to study how adding small representation capacity can already expand its capabilities; (2) investigate how discriminator function capacity and smoothness determine the quality of discovered skills, or latent goals, through modifying latent dimensionality and applying spectral normalization; (3) adapt techniques such as hindsight experience replay (HER) from GCRL to MI-based RL; and lastly, (4) propose a novel evaluation metric, named latent goal reaching (LGR), for comparing empowerment algorithms with different choices of latent dimensionality and discriminator parameterization. Through principled mathematical derivations and careful experimental studies, our work lays a novel foundation from which to evaluate, analyze, and develop representation learning techniques in goal-based RL.

The assessment of program functionality can generally be accomplished with straight-forward unit tests. However, assessing the design quality of a program is a much more difficult and nuanced problem. Design quality is an important consideration since it affects the readability and maintainability of programs. Assessing design quality and giving personalized feedback is very time consuming task for instructors and teaching assistants. This limits the scale of giving personalized feedback to small class settings. Further, design quality is nuanced and is difficult to concisely express as a set of rules. For these reasons, we propose a neural network model to both automatically assess the design of a program and provide personalized feedback to guide students on how to make corrections. The model's effectiveness is evaluated on a corpus of student programs written in Python. The model has an accuracy rate from 83.67% to 94.27%, depending on the dataset, when predicting design scores as compared to historical instructor assessment. Finally, we present a study where students tried to improve the design of their programs based on the personalized feedback produced by the model. Students who participated in the study improved their program design scores by 19.58%.

Abstractive summarization, the task of generating a concise summary of input documents, requires: (1) reasoning over the source document to determine the salient pieces of information scattered across the long document, and (2) composing a cohesive text by reconstructing these salient facts into a shorter summary that faithfully reflects the complex relations connecting these facts. In this paper, we adapt TP-TRANSFORMER (Schlag et al., 2019), an architecture that enriches the original Transformer (Vaswani et al., 2017) with the explicitly compositional Tensor Product Representation (TPR), for the task of abstractive summarization. The key feature of our model is a structural bias that we introduce by encoding two separate representations for each token to represent the syntactic structure (with role vectors) and semantic content (with filler vectors) separately. The model then binds the role and filler vectors into the TPR as the layer output. We argue that the structured intermediate representations enable the model to take better control of the contents (salient facts) and structures (the syntax that connects the facts) when generating the summary. Empirically, we show that our TP-TRANSFORMER outperforms the Transformer and the original TP-TRANSFORMER significantly on several abstractive summarization datasets based on both automatic and human evaluations. On several syntactic and semantic probing tasks, we demonstrate the emergent structural information in the role vectors and improved syntactic interpretability in the TPR layer outputs. Code and models are available at https://github.com/jiangycTarheel/TPT-Summ.

Distributed multi-agent learning enables agents to cooperatively train a model without requiring to share their datasets. While this setting ensures some level of privacy, it has been shown that, even when data is not directly shared, the training process is vulnerable to privacy attacks including data reconstruction and model inversion attacks. Additionally, malicious agents that train on inverted labels or random data, may arbitrarily weaken the accuracy of the global model. This paper addresses these challenges and presents Privacy-preserving and trustable Distributed Learning (PT-DL), a fully decentralized framework that relies on Differential Privacy to guarantee strong privacy protections of the agents' data, and Ethereum smart contracts to ensure trustability. The paper shows that PT-DL is resilient up to a 50% collusion attack, with high probability, in a malicious trust model and the experimental evaluation illustrates the benefits of the proposed model as a privacy-preserving and trustable distributed multi-agent learning system on several classification tasks.