Search

Large neural models have demonstrated human-level performance on language and vision benchmarks, while their performance degrades considerably on adversarial or out-of-distribution samples. This raises the question of whether these models have learned to solve a dataset rather than the underlying task by overfitting to spurious dataset biases. We investigate one recently proposed approach, AFLite, which adversarially filters such dataset biases, as a means to mitigate the prevalent overestimation of machine performance. We provide a theoretical understanding for AFLite, by situating it in the generalized framework for optimum bias reduction. We present extensive supporting evidence that AFLite is broadly applicable for reduction of measurable dataset biases, and that models trained on the filtered datasets yield better generalization to out-of-distribution tasks. Finally, filtering results in a large drop in model performance (e.g., from 92% to 62% for SNLI), while human performance still remains high. Our work thus shows that such filtered datasets can pose new research challenges for robust generalization by serving as upgraded benchmarks., Comment: Accepted to ICML 2020

A group of Early-Career Researchers (ECRs) has been given a mandate from the European Committee for Future Accelerators (ECFA) to debate the topics of the current European Strategy Update (ESU) for Particle Physics and to summarise the outcome in a brief document [1]. A full-day debate with 180 delegates was held at CERN, followed by a survey collecting quantitative input. During the debate, the ECRs discussed future colliders in terms of the physics prospects, their implications for accelerator and detector technology as well as computing and software. The discussion was organised into several topic areas. From these areas two common themes were particularly highlighted by the ECRs: sociological and human aspects; and issues of the environmental impact and sustainability of our research., Comment: Editors: A. Bethani, E. Brondolin, A. A. Elliot, J. Garc\'ia Pardi\~nas, G. Gilles, L. Gouskos, E. Gouveia, E. Graverini, N. Hermansson-Truedsson, A. Irles, H. Jansen, K. H. Mankinen, E. Manoni, A. Mathad, J. McFayden, M. Queitsch-Maitland, J. Rembser, E. T. J. Reynolds, R. Sch\"ofbeck, P. Schwendimann, S. Sekmen, P. Sznajder, S. L. Williams, D. Zanzi

The LUX-ZEPLIN dark matter search aims to achieve a sensitivity to the WIMP-nucleon spin-independent cross-section down to (1--2)$\times10^{-12}$\,pb at a WIMP mass of 40 GeV/$c^2$. This paper describes the simulations framework that, along with radioactivity measurements, was used to support this projection, and also to provide mock data for validating reconstruction and analysis software. Of particular note are the event generators, which allow us to model the background radiation, and the detector response physics used in the production of raw signals, which can be converted into digitized waveforms similar to data from the operational detector. Inclusion of the detector response allows us to process simulated data using the same analysis routines as developed to process the experimental data., Comment: 24 pages, 19 figures; Corresponding Authors: A. Cottle, V. Kudryavtsev, D. Woodward

The LUX-ZEPLIN (LZ) experiment will enable a neutrinoless double beta decay search in parallel to the main science goal of discovering dark matter particle interactions. We report the expected LZ sensitivity to $^{136}$Xe neutrinoless double beta decay, taking advantage of the significant ($>$600 kg) $^{136}$Xe mass contained within the active volume of LZ without isotopic enrichment. After 1000 live-days, the median exclusion sensitivity to the half-life of $^{136}$Xe is projected to be 1.06$\times$10$^{26}$ years (90% confidence level), similar to existing constraints. We also report the expected sensitivity of a possible subsequent dedicated exposure using 90% enrichment with $^{136}$Xe at 1.06$\times$10$^{27}$ years., Comment: 13 pages, 7 figures, 2 tables, version 2 changes: additional clarifications requested by referee on Sections II.A, III.C, III.E, III.F and IV.B

Two-neutrino double electron capture is a process allowed in the Standard Model of Particle Physics. This rare decay has been observed in $^{78}$Kr, $^{130}$Ba and more recently in $^{124}$Xe. In this publication we report on the search for this process in $^{124}$Xe and $^{126}$Xe using the full exposure of the Large Underground Xenon (LUX) experiment, in a total of of 27769.5~kg-days. No evidence of a signal was observed, allowing us to set 90\% C.L. lower limits for the half-lives of these decays of $2.0\times10^{21}$~years for $^{124}$Xe and $1.9\times10^{21}$~years for $^{126}$Xe., Comment: 8 pages, 3 figures

We introduce the notion of pattern for numerical semigroups, which allows us to generalize the definition of Arf numerical semigroups. In this way infinitely many other classes of numerical semigroups are defined giving a classification of the whole set of numerical semigroups. In particular, all semigroups can be arranged in an infinite non-stabilizing ascending chain whose first step consists just of the trivial semigroup and whose second step is the well known class of Arf semigroups. We escribe a procedure to compute the closure of a numerical semigroup with respect to a pattern. By using the concept of system of generators associated to a pattern we construct recursively a directed acyclic graph with all the semigroups admitting the pattern.

To apply eyeshadow without a brush, should I use a cotton swab or a toothpick? Questions requiring this kind of physical commonsense pose a challenge to today's natural language understanding systems. While recent pretrained models (such as BERT) have made progress on question answering over more abstract domains - such as news articles and encyclopedia entries, where text is plentiful - in more physical domains, text is inherently limited due to reporting bias. Can AI systems learn to reliably answer physical common-sense questions without experiencing the physical world? In this paper, we introduce the task of physical commonsense reasoning and a corresponding benchmark dataset Physical Interaction: Question Answering or PIQA. Though humans find the dataset easy (95% accuracy), large pretrained models struggle (77%). We provide analysis about the dimensions of knowledge that existing models lack, which offers significant opportunities for future research., Comment: AAAI 2020

Understanding narratives requires reasoning about implicit world knowledge related to the causes, effects, and states of situations described in text. At the core of this challenge is how to access contextually relevant knowledge on demand and reason over it. In this paper, we present initial studies toward zero-shot commonsense question answering by formulating the task as inference over dynamically generated commonsense knowledge graphs. In contrast to previous studies for knowledge integration that rely on retrieval of existing knowledge from static knowledge graphs, our study requires commonsense knowledge integration where contextually relevant knowledge is often not present in existing knowledge bases. Therefore, we present a novel approach that generates contextually-relevant symbolic knowledge structures on demand using generative neural commonsense knowledge models. Empirical results on two datasets demonstrate the efficacy of our neuro-symbolic approach for dynamically constructing knowledge graphs for reasoning. Our approach achieves significant performance boosts over pretrained language models and vanilla knowledge models, all while providing interpretable reasoning paths for its predictions.

For a numerical semigroup, we encode the set of primitive elements that are larger than its Frobenius number and show how to produce in a fast way the corresponding sets for its children in the semigroup tree. This allows us to present an efficient algorithm for exploring the tree up to a given genus. The algorithm exploits the second nonzero element of a numerical semigroup and the particular pseudo-ordinary case in which this element is the conductor., Comment: Accepted at Mathematics of Computation (AMS)

French recommendations for the screening of hepatitis B virus (HBV) infection were updated in 2019 with the association of three markers: HBs Ag, anti-HBs Ab and anti-HBc Ab. These three markers allow identification of infected patients, vaccinated patients and patients who have been in contact with HBV. A positive HBs Ag is usually associated with HBV infection but this interpretation must take into account the clinical context. In particular, the absence of anti-HBc Ab, normal ALAT levels and the absence of jaundice can be associated with recent HBV vaccination or false-positive HBs Ag. Recent HBV vaccination can usually be confirmed by patient questioning, while confirmatory tests are useful to detect false positive HBs Ag. If necessary, a second sample can be requested to confirm the interpretation., Comment: in French

We describe the design and assembly of the LUX-ZEPLIN experiment, a direct detection search for cosmic WIMP dark matter particles. The centerpiece of the experiment is a large liquid xenon time projection chamber sensitive to low energy nuclear recoils. Rejection of backgrounds is enhanced by a Xe skin veto detector and by a liquid scintillator Outer Detector loaded with gadolinium for efficient neutron capture and tagging. LZ is located in the Davis Cavern at the 4850' level of the Sanford Underground Research Facility in Lead, South Dakota, USA. We describe the major subsystems of the experiment and its key design features and requirements.

We report here methods and techniques for creating and improving a model that reproduces the scintillation and ionization response of a dual-phase liquid and gaseous xenon time-projection chamber. Starting with the recent release of the Noble Element Simulation Technique (NEST v2.0), electronic recoil data from the $\beta$ decays of ${}^3$H and ${}^{14}$C in the Large Underground Xenon (LUX) detector were used to tune the model, in addition to external data sets that allow for extrapolation beyond the LUX data-taking conditions. This paper also presents techniques used for modeling complicated temporal and spatial detector pathologies that can adversely affect data using a simplified model framework. The methods outlined in this report show an example of the robust applications possible with NEST v2.0, while also providing the final electronic recoil model and detector parameters that will used in the new analysis package, the LUX Legacy Analysis Monte Carlo Application (LLAMA), for accurate reproduction of the LUX data. As accurate background reproduction is crucial for the success of rare-event searches, such as dark matter direct detection experiments, the techniques outlined here can be used in other single-phase and dual-phase xenon detectors to assist with accurate ER background reproduction., Comment: 17 Pages, 10 Figures, 2 Tables

Understanding narratives requires reading between the lines, which in turn, requires interpreting the likely causes and effects of events, even when they are not mentioned explicitly. In this paper, we introduce Cosmos QA, a large-scale dataset of 35,600 problems that require commonsense-based reading comprehension, formulated as multiple-choice questions. In stark contrast to most existing reading comprehension datasets where the questions focus on factual and literal understanding of the context paragraph, our dataset focuses on reading between the lines over a diverse collection of people's everyday narratives, asking such questions as "what might be the possible reason of ...?", or "what would have happened if ..." that require reasoning beyond the exact text spans in the context. To establish baseline performances on Cosmos QA, we experiment with several state-of-the-art neural architectures for reading comprehension, and also propose a new architecture that improves over the competitive baselines. Experimental results demonstrate a significant gap between machine (68.4%) and human performance (94%), pointing to avenues for future research on commonsense machine comprehension. Dataset, code and leaderboard is publicly available at https://wilburone.github.io/cosmos., Comment: EMNLP'2019

Abductive reasoning is inference to the most plausible explanation. For example, if Jenny finds her house in a mess when she returns from work, and remembers that she left a window open, she can hypothesize that a thief broke into her house and caused the mess, as the most plausible explanation. While abduction has long been considered to be at the core of how people interpret and read between the lines in natural language (Hobbs et al., 1988), there has been relatively little research in support of abductive natural language inference and generation. We present the first study that investigates the viability of language-based abductive reasoning. We introduce a challenge dataset, ART, that consists of over 20k commonsense narrative contexts and 200k explanations. Based on this dataset, we conceptualize two new tasks -- (i) Abductive NLI: a multiple-choice question answering task for choosing the more likely explanation, and (ii) Abductive NLG: a conditional generation task for explaining given observations in natural language. On Abductive NLI, the best model achieves 68.9% accuracy, well below human performance of 91.4%. On Abductive NLG, the current best language generators struggle even more, as they lack reasoning capabilities that are trivial for humans. Our analysis leads to new insights into the types of reasoning that deep pre-trained language models fail to perform--despite their strong performance on the related but more narrowly defined task of entailment NLI--pointing to interesting avenues for future research., Comment: ICLR 2020 Camera Ready

We present the results of a direct detection search for mirror dark matter interactions, using data collected from the Large Underground Xenon experiment during 2013, with an exposure of 95 live-days $\times$ 118 kg. Here, the calculations of the mirror electron scattering rate in liquid xenon take into account the shielding effects from mirror dark matter captured within the Earth. Annual and diurnal modulation of the dark matter flux and atomic shell effects in xenon are also accounted for. Having found no evidence for an electron recoil signal induced by mirror dark matter interactions we place an upper limit on the kinetic mixing parameter over a range of local mirror electron temperatures between 0.1 and 0.6 keV. This limit shows significant improvement over the previous experimental constraint from orthopositronium decays and significantly reduces the allowed parameter space for the model. We exclude mirror electron temperatures above 0.3 keV at a 90% confidence level, for this model, and constrain the kinetic mixing below this temperature.

The Winograd Schema Challenge (WSC) (Levesque, Davis, and Morgenstern 2011), a benchmark for commonsense reasoning, is a set of 273 expert-crafted pronoun resolution problems originally designed to be unsolvable for statistical models that rely on selectional preferences or word associations. However, recent advances in neural language models have already reached around 90% accuracy on variants of WSC. This raises an important question whether these models have truly acquired robust commonsense capabilities or whether they rely on spurious biases in the datasets that lead to an overestimation of the true capabilities of machine commonsense. To investigate this question, we introduce WinoGrande, a large-scale dataset of 44k problems, inspired by the original WSC design, but adjusted to improve both the scale and the hardness of the dataset. The key steps of the dataset construction consist of (1) a carefully designed crowdsourcing procedure, followed by (2) systematic bias reduction using a novel AfLite algorithm that generalizes human-detectable word associations to machine-detectable embedding associations. The best state-of-the-art methods on WinoGrande achieve 59.4-79.1%, which are 15-35% below human performance of 94.0%, depending on the amount of the training data allowed. Furthermore, we establish new state-of-the-art results on five related benchmarks - WSC (90.1%), DPR (93.1%), COPA (90.6%), KnowRef (85.6%), and Winogender (97.1%). These results have dual implications: on one hand, they demonstrate the effectiveness of WinoGrande when used as a resource for transfer learning. On the other hand, they raise a concern that we are likely to be overestimating the true capabilities of machine commonsense across all these benchmarks. We emphasize the importance of algorithmic bias reduction in existing and future benchmarks to mitigate such overestimation.

We prove that for a quasi-regular semiperfectoid $\mathbb{Z}_p^{\rm cycl}$-algebra $R$ (in the sense of Bhatt-Morrow-Scholze), the cyclotomic trace map from the $p$-completed $K$-theory spectrum $K(R;\mathbb{Z}_p)$ of $R$ to the topological cyclic homology $\mathrm{TC}(R;\mathbb{Z}_p)$ of $R$ identifies on $\pi_2$ with a $q$-deformation of the logarithm.

We define, for each quasi-syntomic ring $R$ (in the sense of Bhatt-Morrow-Scholze), a category $\mathrm{DM}^{\rm adm}(R)$ of \textit{admissible prismatic Dieudonn\'e crystals over $R$} and a natural functor from $p$-divisible groups over $R$ to $\mathrm{DM}^{\rm adm}(R)$. We prove that this functor is an antiequivalence. Our main cohomological tool is the prismatic formalism recently developed by Bhatt and Scholze., Comment: Replaced the notion of filtered prismatic Dieudonn\'e crystal by the equivalent (but simpler) notion of admissible prismatic Dieudonn\'e crystal; corrected and simplified (following a suggestion of Mathew) the proof of fully faithfulness; many small changes

We present a novel analysis technique for liquid xenon time projection chambers that allows for a lower threshold by relying on events with a prompt scintillation signal consisting of single detected photons. The energy threshold of the LUX dark matter experiment is primarily determined by the smallest scintillation response detectable, which previously required a 2-fold coincidence signal in its photomultiplier arrays, enforced in data analysis. The technique presented here exploits the double photoelectron emission effect observed in some photomultiplier models at vacuum ultraviolet wavelengths. We demonstrate this analysis using an electron recoil calibration dataset and place new constraints on the spin-independent scattering cross section of weakly interacting massive particles (WIMPs) down to 2.5 GeV/c$^2$ WIMP mass using the 2013 LUX dataset. This new technique is promising to enhance light WIMP and astrophysical neutrino searches in next-generation liquid xenon experiments.

Consider a complete flag $\{0\} = C_0 < C_1 < \cdots < C_n = \mathbb{F}^n$ of one-point AG codes of length $n$ over the finite field $\mathbb{F}$. The codes are defined by evaluating functions with poles at a given point $Q$ in points $P_1,\dots,P_n$ distinct from $Q$. A flag has the isometry-dual property if the given flag and the corresponding dual flag are the same up to isometry. For several curves, including the projective line, Hermitian curves, Suzuki curves, Ree curves, and the Klein curve over the field of eight elements, the maximal flag, obtained by evaluation in all rational points different from the point $Q$, is self-dual. More generally, we ask whether a flag obtained by evaluation in a proper subset of rational points is isometry-dual. In [3] it is shown, for a curve of genus $g$, that a flag of one-point AG codes defined with a subset of $n > 2g+2$ rational points is isometry-dual if and only if the last code $C_n$ in the flag is defined with functions of pole order at most $n+2g-1$. Using a different approach, we extend this characterization to all subsets of size $n \geq 2g+2$. Moreover we show that this is best possible by giving examples of isometry-dual flags with $n=2g+1$ such that $C_n$ is generated by functions of pole order at most $n+2g-2$. We also prove a necessary condition, formulated in terms of maximum sparse ideals of the Weierstrass semigroup of $Q$, under which a flag of punctured one-point AG codes inherits the isometry-dual property from the original unpunctured flag., Comment: 25 pages

Collaboration between human supervisors and remote teams of robots is highly challenging, particularly in high-stakes, distant, hazardous locations, such as off-shore energy platforms. In order for these teams of robots to truly be beneficial, they need to be trusted to operate autonomously, performing tasks such as inspection and emergency response, thus reducing the number of personnel placed in harm's way. As remote robots are generally trusted less than robots in close-proximity, we present a solution to instil trust in the operator through a `mediator robot' that can exhibit social skills, alongside sophisticated visualisation techniques. In this position paper, we present general challenges and then take a closer look at one challenge in particular, discussing an initial study, which investigates the relationship between the level of control the supervisor hands over to the mediator robot and how this affects their trust. We show that the supervisor is more likely to have higher trust overall if their initial experience involves handing over control of the emergency situation to the robotic assistant. We discuss this result, here, as well as other challenges and interaction techniques for human-robot collaboration., Comment: 9 pages. Peer reviewed position paper accepted in the CHI 2019 Workshop: The Challenges of Working on Social Robots that Collaborate with People (SIRCHI2019), ACM CHI Conference on Human Factors in Computing Systems, May 2019, Glasgow, UK

We analyze the set of increasingly enumerable additive submonoids of R, for instance, the set of logarithms of the positive integers with respect to a given base. We call them $\omega$-monoids. The $\omega$-monoids for which consecutive elements become arbitrarily close are called tempered monoids. This is, in particular, the case for the set of logarithms. We show that any $\omega$-monoid is either a scalar multiple of a numerical semigroup or a tempered monoid. We will also show how we can differentiate $\omega$-monoids that are multiples of numerical semigroups from those that are tempered monoids by the size and commensurability of their minimal generating sets. All the definitions and results are illustrated with examples from music theory., Comment: Accepted at American Mathematical Monthly. This paper is the second part of arXiv:1703.01077

Deep underground environments are ideal for low background searches due to the attenuation of cosmic rays by passage through the earth. However, they are affected by backgrounds from $\gamma$-rays emitted by $^{40}$K and the $^{238}$U and $^{232}$Th decay chains in the surrounding rock. The LUX-ZEPLIN (LZ) experiment will search for dark matter particle interactions with a liquid xenon TPC located within the Davis campus at the Sanford Underground Research Facility, Lead, South Dakota, at the 4,850-foot level. In order to characterise the cavern background, in-situ $\gamma$-ray measurements were taken with a sodium iodide detector in various locations and with lead shielding. The integral count rates (0--3300~keV) varied from 596~Hz to 1355~Hz for unshielded measurements, corresponding to a total flux in the cavern of $1.9\pm0.4$~$\gamma~$cm$^{-2}$s$^{-1}$. The resulting activity in the walls of the cavern can be characterised as $220\pm60$~Bq/kg of $^{40}$K, $29\pm15$~Bq/kg of $^{238}$U, and $13\pm3$~Bq/kg of $^{232}$Th., Comment: 11 pages, 9 figures

Ultra-SAXS can enhance the capabilities of existing SAXS/WAXS beamlines and laboratory instruments. A compact Ultra-SAXS module has been developed, which extends the measurable $q$-range with $0.0015 \leq q\mathrm{(nm^{-1})} \leq 0.2$, allowing structural dimensions between $30 \leq D$(nm)$ \leq 4000$ to be probed in addition to the range covered by a high-end SAXS/WAXS instrument. By shifting the module components in and out on their respective motor stages, SAXS/WAXS measurements can be easily and rapidly interleaved with USAXS measurements. In this paper, the design considerations, realization and synchrotron findings are presented. Measurements of silica spheres, an alumina membrane, and a porous carbon catalyst are provided as application examples.

We report results from an extensive set of measurements of the \b{eta}-decay response in liquid xenon.These measurements are derived from high-statistics calibration data from injected sources of both $^{3}$H and $^{14}$C in the LUX detector. The mean light-to-charge ratio is reported for 13 electric field values ranging from 43 to 491 V/cm, and for energies ranging from 1.5 to 145 keV.

In this work, we experimentally assess the transmission of a PDCP-RLC virtualised RAN split interface through a commercial XGS-PON system. We investigate the impacts of DBA on the uplink and packet jitter on the downlink.

We successfully demonstrate a transmission of a high layer split mobile interface for cell-less, high-speed train network applications using a commercially available XGS-PON. Operation is also demonstrated for a GbE interface.

In this work the development results of the TRITIUM project is presented. The main objective of the project is the construction of a near real-time monitor for low activity tritium in water, aimed at in-situ surveillance and radiological protection of river water in the vicinity of nuclear power plants. The European Council Directive 2013/51/Euratom requires that the maximum level of tritium in water for human consumption to be lower than 100 Bq/L. Tritium levels in the cooling water of nuclear power plants in normal operation are much higher than the levels caused by the natural and cosmogenic components, and may easily surmount the limit required by the Directive. The current liquid-scintillation measuring systems in environmental radioactivity laboratories are sensitive to such low levels, but they are not suitable for real-time monitoring. Moreover, there is no currently available device with enough sensitivity and monitoring capabilities that could be used for surveillance of the cooling water of nuclear power plants. A detector system based on scintillation fibers read out by photomultiplier tubes (PMTs) or silicon photomultiplier (SiPM) arrays is under development for in-water tritium measurement. This detector will be installed in the vicinity of Almaraz nuclear power plant (Spain) in Spring 2019. An overview of the project development and the results of first prototypes are presented., Comment: 6 pages, 11 figures, IEEE NSS-MIC Conf Record 2018