Search for: All records

Elastocapillary rolling transfer weaves soft materials to spatial structures for programmable robotic applications. 

Sterile neutrinos only interact with the standard model through the neutrino sector, and thus represent a simple dark matter (DM) candidate with many potential astrophysical and cosmological signatures. Recently, sterile neutrinos produced through self-interactions of active neutrinos have received attention as a particle candidate that can yield the entire observed DM relic abundance without violating the most stringent constraints from X-ray observations. We examine consistency of this production mechanism with the abundance of small-scale structure in the universe, as captured by the population of ultrafaint dwarf galaxies orbiting the Milky Way, and derive a lower bound on the sterile-neutrino particle mass of 37 keV. Combining these results with previous collider and X-ray limits excludes 100% sterile-neutrino DM produced by strong neutrino self-coupling, mediated by a heavy (≳1 GeV) scalar; however, data permits sterile-neutrino DM production via a light mediator.

 

Language understanding is essential for the navigation agent to follow instructions. We observe two kinds of issues in the instructions that can make the navigation task challenging: 1. The mentioned landmarks are not recognizable by the navigation agent due to the different vision abilities of the instructor and the modeled agent. 2. The mentioned landmarks are applicable to multiple targets, thus not distinctive for selecting the target among the candidate viewpoints. To deal with these issues, we design a translator module for the navigation agent to convert the original instructions into easy-to-follow sub-instruction representations at each step. The translator needs to focus on the recognizable and distinctive landmarks based on the agent’s visual abilities and the observed visual environment. To achieve this goal, we create a new synthetic sub-instruction dataset and design specific tasks to train the translator and the navigation agent. We evaluate our approach on Room2Room (R2R), Room4room (R4R), and Room2Room Last (R2R-Last) datasets and achieve state-of-the-art results on multiple benchmarks. 

Recent research has shown that integrating domain knowledge into deep learning architectures is effective – it helps reduce the amount of required data, improves the accuracy of the models’ decisions, and improves the interpretability of models. However, the research community is missing a convened benchmark for systematically evaluating knowledge integration methods. In this work, we create a benchmark that is a collection of nine tasks in the domains of natural language processing and computer vision. In all cases, we model external knowledge as constraints, specify the sources of the constraints for each task, and implement various models that use these constraints. We report the results of these models using a new set of extended evaluation criteria in addition to the task performances for a more in-depth analysis. This effort provides a framework for a more comprehensive and systematic comparison of constraint integration techniques and for identifying related research challenges. It will facilitate further research for alleviating some problems of state-of-the-art neural models. 

The practice of infrastructure as code (IaC) recommends automated management of computing infrastructure with application of quality assurance, such as linting and testing. To that end, researchers recently have investigated quality concerns in IaC test manifests by deriving a catalog of test smells. The relevance of the identified smells need to be quantified by obtaining feedback from practitioners. Such feedback can help the IaC community understand if smells have relevance amongst practitioners, and derive future research directions. We survey 30 practitioners to assess the relevance of three Ansible test smell categories namely, assertion roulette, local only testing, and remote mystery guest. We observe local only testing to be the most agreed upon test smell category, whereas, assertion roulette is the least agreed upon test smell category. Our findings provide a nuanced perspective of test smells for IaC, and lays the groundwork for future research. 

Infrastructure as code (IaC) is the practice of automatically managing computing infrastructure at scale. Despite yielding multiple benefits for organizations, the practice of IaC is susceptible to quality concerns, which can lead to large-scale consequences. While researchers have studied quality concerns in IaC manifests, quality aspects of infrastructure orchestrators, i.e., tools that implement the practice of IaC, remain an under-explored area. A systematic investigation of defects in infrastructure orchestrators can help foster further research in the domain of IaC. From our empirical study with 22,445 commits mined from the Ansible infrastructure orchestrator we observe (i) a defect density of 17.9 per KLOC, (ii) 12 categories of Ansible components for which defects appear, and (iii) the ‘Module’ component to include more defects than the other 11 components. Based on our empirical study, we provide recommendations for researchers to conduct future research to enhance the quality of infrastructure orchestrators.