Search for: All records

Transformers have revolutionized machine learning, yet their inner workings remain opaque to many. We present TRANSFORMER EXPLAINER, an interactive visualization tool designed for non-experts to learn about Transformers through the GPT-2 model. Our tool helps users understand complex Transformer concepts by integrating a model overview and smooth transitions across abstraction levels of math operations and model structures. It runs a live GPT-2 model locally in the user’s browser, empowering users to experiment with their own input and observe in real-time how the internal components and parameters of the Transformer work together to predict the next tokens. 125,000 users have used our open-source tool at https://poloclub.github.io/ transformer-explainer/. 

Retrieving evidence to support or refute claims is a core part of automatic fact-checking. Prior work makes simplifying assumptions in retrieval that depart from real-world use cases: either no access to evidence, access to evidence curated by a human fact-checker, or access to evidence published after a claim was made. In this work, we present the first realistic pipeline to check real-world claims by retrieving raw evidence from the web. We restrict our retriever to only search documents available prior to the claim’s making, modeling the realistic scenario of emerging claims. Our pipeline includes five components: claim decomposition, raw document retrieval, fine-grained evidence retrieval, claim-focused summarization, and veracity judgment. We conduct experiments on complex political claims in the ClaimDecomp dataset and show that the aggregated evidence produced by our pipeline improves veracity judgments. Human evaluation finds the evidence summary produced by our system is reliable (it does not hallucinate information) and relevant to answering key questions about a claim, suggesting that it can assist fact-checkers even when it does not reflect a complete evidence set. 

Worked examples are an educational tool widely used in introductory computer science classes, primarily for programming and code-tracing concepts. Prior research supports the use of worked examples as a scaffolding mechanism to help students build a solid foundation before tackling problems on their own. Whether breaking down the intricacies of code or explaining abstract theoretical concepts, worked examples offer a structured approach that nurtures a deeper understanding during self-study. This study explores how peer-created worked examples, shown through detailed step-by-step videos, aid student learning in an intermediate-level computer science course, namely computer systems. Our results suggest that worked-example videos are a useful study aid for intermediate computer science courses, such as computer systems. Students who watched the worked-example videos found them to be very helpful, and ranked them as the top study aid for succeeding on quizzes. Additionally, students with access to worked-example videos performed moderately better on quizzes compared to students without worked-example videos. Our results and experiences also suggest that worked-example videos are beneficial to the students who created them as well as their peers who use them. 

Stroke commonly affects the ability of the upper extremities (UEs) to move normally. In clinical settings, identifying and measuring movement abnormality is challenging due to the imprecision and impracticality of available assessments. These challenges interfere with therapeutic tracking, communication, and treatment. We thus sought to develop an approach that blends precision and pragmatism, combining high-dimensional motion capture with out-of-distribution (OOD) detection. We used an array of wearable inertial measurement units to capture upper body motion in healthy and chronic stroke subjects performing a semi-structured, unconstrained 3D tabletop task. After data were labeled by human coders, we trained two deep learning models exclusively on healthy subject data to classify elemental movements (functional primitives). We tested these healthy subject-trained models on previously unseen healthy and stroke motion data. We found that model confidence, indexed by prediction probabilities, was generally high for healthy test data but significantly dropped when encountering OOD stroke data. Prediction probabilities worsened with more severe motor impairment categories and were directly correlated with individual impairment scores. Data inputs from the paretic UE, rather than trunk, most strongly influenced model confidence. We demonstrate for the first time that using OOD detection with high-dimensional motion data can reveal clinically meaningful movement abnormality in subjects with chronic stroke. 

The future of improved immunotherapy against cancer depends on an in-depth understanding of the dynamic interactions between the immune system and tumors. Over the past two decades, the zebrafish has served as a valuable model system to provide fresh insights into both the development of the immune system and the etiologies of many different cancers. This well-established foundation of knowledge combined with the imaging and genetic capacities of the zebrafish provides a new frontier in cancer immunology research. In this review, we provide an overview of the development of the zebrafish immune system along with a side-by-side comparison of its human counterpart. We then introduce components of the adaptive immune system with a focus on their roles in the tumor microenvironment (TME) of teleosts. In addition, we summarize zebrafish models developed for the study of cancer and adaptive immunity along with other available tools and technology afforded by this experimental system. Finally, we discuss some recent research conducted using the zebrafish to investigate adaptive immune cell-tumor interactions. Without a doubt, the zebrafish will arise as one of the driving forces to help expand the knowledge of tumor immunity and facilitate the development of improved anti-cancer immunotherapy in the foreseeable future. 

Abstract The merger of a binary black hole gives birth to a highly distorted final black hole. The gravitational radiation emitted as this black hole relaxes presents us with the unique opportunity to probe extreme gravity and its connection with the dynamics of the black hole horizon. Using numerical relativity simulations, we demonstrate a connection between a concrete observable feature in the gravitational waves and geometrical features on the dynamical apparent horizon of the final black hole. Specifically, we show how the line-of-sight passage of a “cusp”-like defect on the horizon of the final black hole correlates with “chirp”-like frequency peaks in the post-merger gravitational-waves. These post-merger chirps should be observed and analyzed as the sensitivity of LIGO and Virgo increase and as future generation detectors, such as LISA and the Einstein Telescope, become operational. 

Abstract BackgroundIdiopathic pulmonary fibrosis (IPF) is a progressive, irreversible, and usually fatal lung disease of unknown reasons, generally affecting the elderly population. Early diagnosis of IPF is crucial for triaging patients’ treatment planning into anti‐fibrotic treatment or treatments for other causes of pulmonary fibrosis. However, current IPF diagnosis workflow is complicated and time‐consuming, which involves collaborative efforts from radiologists, pathologists, and clinicians and it is largely subject to inter‐observer variability. PurposeThe purpose of this work is to develop a deep learning‐based automated system that can diagnose subjects with IPF among subjects with interstitial lung disease (ILD) using an axial chest computed tomography (CT) scan. This work can potentially enable timely diagnosis decisions and reduce inter‐observer variability. MethodsOur dataset contains CT scans from 349 IPF patients and 529 non‐IPF ILD patients. We used 80% of the dataset for training and validation purposes and 20% as the holdout test set. We proposed a two‐stage model: at stage one, we built a multi‐scale, domain knowledge‐guided attention model (MSGA) that encouraged the model to focus on specific areas of interest to enhance model explainability, including both high‐ and medium‐resolution attentions; at stage two, we collected the output from MSGA and constructed a random forest (RF) classifier for patient‐level diagnosis, to further boost model accuracy. RF classifier is utilized as a final decision stage since it is interpretable, computationally fast, and can handle correlated variables. Model utility was examined by (1) accuracy, represented by the area under the receiver operating characteristic curve (AUC) with standard deviation (SD), and (2) explainability, illustrated by the visual examination of the estimated attention maps which showed the important areas for model diagnostics. ResultsDuring the training and validation stage, we observe that when we provide no guidance from domain knowledge, the IPF diagnosis model reaches acceptable performance (AUC±SD = 0.93±0.07), but lacks explainability; when including only guided high‐ or medium‐resolution attention, the learned attention maps are not satisfactory; when including both high‐ and medium‐resolution attention, under certain hyperparameter settings, the model reaches the highest AUC among all experiments (AUC±SD = 0.99±0.01) and the estimated attention maps concentrate on the regions of interests for this task. Three best‐performing hyperparameter selections according to MSGA were applied to the holdout test set and reached comparable model performance to that of the validation set. ConclusionsOur results suggest that, for a task with only scan‐level labels available, MSGA+RF can utilize the population‐level domain knowledge to guide the training of the network, which increases both model accuracy and explainability.