Search

While neutrinos are often treated as a background for many dark matter experiments, these particles offer a new avenue for physics: the detection of core-collapse supernovae. Supernovae are extremely energetic, violent and complex events that mark the death of massive stars. During their collapse stars emit a large number of neutrinos in a short burst. These neutrinos carry 99\% of the emitted energy which makes their detection fundamental in understanding supernovae. This paper illustrates how COSINUS (Cryogenic Observatory for SIgnatures seen in Next-generation Underground Searches), a sodium iodide (NaI) based dark matter search, will be sensitive to the next galactic core-collapse supernova. The experiment is composed of two separate detectors which will be sensitive to far and nearby supernovae. The inner core of the experiment will consist of NaI crystals operating as scintillating calorimeters, mainly sensitive to the Coherent Elastic Scattering of Neutrinos (CE$\nu$NS) against the Na and I nuclei. The low mass of the cryogenic detectors gives the experiment a sensitivity to close supernovae below 1kpc without pileup. They will see up to hundreds of CE$\nu$NS events from a supernova happening at 200pc. The crystals reside at the center of a cylindrical 230T water tank, instrumented with 30 photomultipliers. This tank acts as a passive and active shield able to detect the Cherenkov radiation induced by impinging charged particles from ambient and cosmogenic radioactivity. A supernova near the Milky Way Center (10kpc) will be easily detected inducing $\sim$60 measurable events, and the water tank will have a 3$\sigma$ sensitivity to supernovae up to 22kpc, seeing $\sim$10 events. This paper shows how, even without dedicated optimization, modern dark matter experiments will also play their part in the multi-messenger effort to detect the next galactic core-collapse supernova.

The main goal of the CRESST-III experiment is the direct detection of dark matter particles via their scattering off target nuclei in cryogenic detectors. In this work we present the results of a Silicon-On-Sapphire (SOS) detector with a mass of 0.6$\,$g and an energy threshold of (6.7$\, \pm \,$0.2)$\,$eV with a baseline energy resolution of (1.0$\, \pm \,$0.2)$\,$eV. This allowed for a calibration via the detection of single luminescence photons in the eV-range, which could be observed in CRESST for the first time. We present new exclusion limits on the spin-independent and spin-dependent dark matter-nucleon cross section that extend to dark matter particle masses of less than 100$\,$MeV/c$^{2}$., Comment: 10 pages, 8 figures

COSINUS is a dark matter (DM) direct search experiment that uses sodium iodide (NaI) crystals as cryogenic calorimeters. Thanks to the low nuclear recoil energy threshold and event-by-event discrimination capability, COSINUS will address the long-standing DM claim made by the DAMA/LIBRA collaboration. The experiment is currently under construction at the Laboratori Nazionali del Gran Sasso, Italy, and employs a large cylindrical water tank as a passive shield to meet the required background rate. However, muon-induced neutrons can mimic a DM signal therefore requiring an active veto system, which is achieved by instrumenting the water tank with an array of photomultiplier tubes (PMTs). This study optimizes the number, arrangement, and trigger conditions of the PMTs as well as the size of an optically invisible region. The objective was to maximize the muon veto efficiency while minimizing the accidental trigger rate due to the ambient and instrumental background. The final configuration predicts a veto efficiency of 99.63 $\pm$ 0.16 $\%$ and 44.4 $\pm$ $5.6\%$ in the tagging of muon events and showers of secondary particles, respectively. The active veto will reduce the cosmogenic neutron background rate to 0.11 $\pm$ 0.02 cts$\cdot$kg$^{-1}$$\cdot$year$^{-1}$, corresponding to less than one background event in the region of interest for the whole COSINUS-1$\pi$ exposure of 1000 kg$\cdot$days.

In recent times, the sensitivity of low-mass direct dark matter searches has been limited by unknown low energy backgrounds close to the energy threshold of the experiments known as the low energy excess (LEE). The CRESST experiment utilises advanced cryogenic detectors constructed with different types of crystals equipped with Transition Edge Sensors (TESs) to measure signals of nuclear recoils induced by the scattering of dark matter particles in the detector. In CRESST, this low energy background manifests itself as a steeply rising population of events below 200 eV. A novel detector design named doubleTES using two identical TESs on the target crystal was studied to investigate the hypothesis that the events are sensor-related. We present the first results from two such modules, demonstrating their ability to differentiate between events originating from the crystal's bulk and those occurring in the sensor or in its close proximity., Comment: 10 pages, 13 figures

GNN-based approaches for learning general policies across planning domains are limited by the expressive power of $C_2$, namely; first-order logic with two variables and counting. This limitation can be overcomed by transitioning to $k$-GNNs, for $k=3$, wherein object embeddings are substituted with triplet embeddings. Yet, while $3$-GNNs have the expressive power of $C_3$, unlike $1$- and $2$-GNNs that are confined to $C_2$, they require quartic time for message exchange and cubic space for embeddings, rendering them impractical. In this work, we introduce a parameterized version of relational GNNs. When $t$ is infinity, R-GNN[$t$] approximates $3$-GNNs using only quadratic space for embeddings. For lower values of $t$, such as $t=1$ and $t=2$, R-GNN[$t$] achieves a weaker approximation by exchanging fewer messages, yet interestingly, often yield the $C_3$ features required in several planning domains. Furthermore, the new R-GNN[$t$] architecture is the original R-GNN architecture with a suitable transformation applied to the input states only. Experimental results illustrate the clear performance gains of R-GNN[$1$] and R-GNN[$2$] over plain R-GNNs, and also over edge transformers that also approximate $3$-GNNs., Comment: Submitted to IJCAI 2024

Cryogenic scintillating calorimeters are ultrasensitive particle detectors for rare event searches, particularly for the search for dark matter and the measurement of neutrino properties. These detectors are made from scintillating target crystals generating two signals for each particle interaction. The phonon (heat) signal precisely measures the deposited energy independent of the type of interacting particle. The scintillation light signal yields particle discrimination on an event-by-event basis. This paper presents a likelihood framework modeling backgrounds and a potential dark matter signal in the two-dimensional plane spanned by phonon and scintillation light energies. We apply the framework to data from CaWO$_4$-based detectors operated in the CRESST dark matter search. For the first time, a single likelihood framework is used in CRESST to model the data and extract results on dark matter in one step by using a profile likelihood ratio test. Our framework simultaneously fits (neutron) calibration data and physics (background) data and allows combining data from multiple detectors. Although tailored to CaWO$_4$-targets and the CRESST experiment, the framework can easily be expanded to other materials and experiments using scintillating cryogenic calorimeters for dark matter search and neutrino physics., Comment: 18 pages, 11 figures, additional figures and data in ancillary files (on arXiv or at journal reference)

The CSC (cryogenic scintillating calorimeter) technology devoted to rare event searches is reaching the sensitivity level required for the hunt of dark matter-electron scatterings. Dark matter-electron interactions in scintillating targets are expected to stimulate the emission of single photons, each of energy equal to the target electronic band gap. The electronic band gap in scintillators like NaI/GaAs is of O(eV). The search for this signal can be done by an array of cryogenic light detectors with eV/sub-eV energy resolution. In this work, we describe the detection principle, the detector response and the envisioned detector design to search for dark matter interacting with electrons via the measurement of the scintillation light at millikelvin. First sensitivity projections are provided, which show the potential of this research.

Recently low-mass dark matter direct searches have been hindered by a low energy background, drastically reducing the physics reach of the experiments. In the CRESST-III experiment, this signal is characterised by a significant increase of events below 200 eV. As the origin of this background is still unknown, it became necessary to develop new detector designs to reach a better understanding of the observations. Within the CRESST collaboration, three new different detector layouts have been developed and they are presented in this contribution., Comment: 8 pages, 4 figures

In the impartial selection problem, a subset of agents up to a fixed size $k$ among a group of $n$ is to be chosen based on votes cast by the agents themselves. A selection mechanism is impartial if no agent can influence its own chance of being selected by changing its vote. It is $\alpha$-optimal if, for every instance, the ratio between the votes received by the selected subset is at least a fraction of $\alpha$ of the votes received by the subset of size $k$ with the highest number of votes. We study deterministic impartial mechanisms in a more general setting with arbitrarily weighted votes and provide the first approximation guarantee, roughly $1/\lceil 2n/k\rceil$. When the number of agents to select is large enough compared to the total number of agents, this yields an improvement on the previously best known approximation ratio of $1/k$ for the unweighted setting. We further show that our mechanism can be adapted to the impartial assignment problem, in which multiple sets of up to $k$ agents are to be selected, with a loss in the approximation ratio of $1/2$., Comment: Extended version to appear in the ACM Transactions on Economics and Computation (TEAC)

One challenge in speech translation is that plenty of spoken content is long-form, but short units are necessary for obtaining high-quality translations. To address this mismatch, we adapt large language models (LLMs) to split long ASR transcripts into segments that can be independently translated so as to maximize the overall translation quality. We overcome the tendency of hallucination in LLMs by incorporating finite-state constraints during decoding; these eliminate invalid outputs without requiring additional training. We discover that LLMs are adaptable to transcripts containing ASR errors through prompt-tuning or fine-tuning. Relative to a state-of-the-art automatic punctuation baseline, our best LLM improves the average BLEU by 2.9 points for English-German, English-Spanish, and English-Arabic TED talk translation in 9 test sets, just by improving segmentation., Comment: accepted to the Findings of EMNLP 2023. arXiv admin note: text overlap with arXiv:2212.09895

Diamond operated as a cryogenic calorimeter is an excellent target for direct detection of low-mass dark matter candidates. Following the realization of the first low-threshold cryogenic detector that uses diamond as absorber for astroparticle physics applications, we now present the resulting exclusion limits on the elastic spin-independent interaction cross-section of dark matter with diamond. We measured two 0.175 g CVD (Chemical Vapor Deposition) diamond samples, each instrumented with a W-TES. Thanks to the energy threshold of just 16.8 eV of one of the two detectors, we set exclusion limits on the elastic spin-independent interaction of dark matter particles with carbon nuclei down to dark matter masses as low as 0.122 GeV/c2. This work shows the scientific potential of cryogenic detectors made from diamond and lays the foundation for the use of this material as target for direct detection dark matter experiments., Comment: 6 pages, 6 figures

Large Language Models (LLMs) have demonstrated impressive capabilities for text rewriting. Nonetheless, the large sizes of these models make them impractical for on-device inference, which would otherwise allow for enhanced privacy and economical inference. Creating a smaller yet potent language model for text rewriting presents a formidable challenge because it requires balancing the need for a small size with the need to retain the emergent capabilities of the LLM, that requires costly data collection. To address the above challenge, we introduce a new instruction tuning approach for building a mobile-centric text rewriting model. Our strategies enable the generation of high quality training data without any human labeling. In addition, we propose a heuristic reinforcement learning framework which substantially enhances performance without requiring preference data. To further bridge the performance gap with the larger server-side model, we propose an effective approach that combines the mobile rewrite agent with the server model using a cascade. To tailor the text rewriting tasks to mobile scenarios, we introduce MessageRewriteEval, a benchmark that focuses on text rewriting for messages through natural language instructions. Through empirical experiments, we demonstrate that our on-device model surpasses the current state-of-the-art LLMs in text rewriting while maintaining a significantly reduced model size. Notably, we show that our proposed cascading approach improves model performance.

Sodium iodide (NaI) based cryogenic scintillating calorimeters using quantum sensors for signal read out have shown promising first results towards a model-independent test of the annually modulating signal detected by the DAMA/LIBRA dark matter experiment. The COSINUS collaboration has previously reported on the first above-ground measurements using a dual channel readout of phonons and light based on transition edge sensors (TESs) that allows for particle discrimination on an event-by-event basis. In this letter, we outline the first underground measurement of a NaI cryogenic calorimeter read out via the novel remoTES scheme. A 3.67 g NaI absorber with an improved silicon light detector design was operated at the Laboratori Nazionali del Gran Sasso, Italy. A significant improvement in the discrimination power of $e^-$/$\gamma$-events to nuclear recoils was observed with a five-fold improvement in the nuclear recoil baseline resolution, achieving $\sigma$ = 441 eV. Furthermore, we present a limit on the spin-independent dark-matter nucleon elastic scattering cross-section achieving a sensitivity of $\mathcal{O}$(pb) with an exposure of only 11.6 g d., Comment: 11 pages, 14 figures

The COSINUS direct dark matter experiment situated at Laboratori Nazionali del Gran Sasso in Italy is set to investigate the nature of the annually modulating signal detected by the DAMA/LIBRA experiment. COSINUS has already demonstrated that sodium iodide crystals can be operated at mK temperature as cryogenic scintillating calorimeters using transition edge sensors, despite the complication of handling a hygroscopic and low melting point material. With results from a new COSINUS prototype, we show that particle discrimination on an event-by-event basis in NaI is feasible using the dual-channel readout of both phonons and scintillation light. The detector was mounted in the novel remoTES design and operated in an above-ground facility for 9.06 g$\cdot$d of exposure. With a 3.7 g NaI crystal, e$^-$/$\gamma$ events could be clearly distinguished from nuclear recoils down to the nuclear recoil energy threshold of 15 keV., Comment: 7 pages, 9 figures

Using CaWO$_4$ crystals as cryogenic calorimeters, the CRESST experiment searches for nuclear recoils caused by the scattering of potential Dark Matter particles. A reliable identification of a potential signal crucially depends on an accurate background model. In this work we introduce an improved normalisation method for CRESST's model of the electromagnetic backgrounds. Spectral templates, based on Geant4 simulations, are normalised via a Bayesian likelihood fit to experimental background data. Contrary to our previous work, no assumption of partial secular equilibrium is required, which results in a more robust and versatile applicability. Furthermore, considering the correlation between all background components allows us to explain 82.7% of the experimental background within [1 keV, 40 keV], an improvement of 18.6% compared to our previous method., Comment: 24 pages, 14 figures, submitted to EPJC

New-generation direct searches for low mass dark matter feature detection thresholds at energies well below 100 eV, much lower than the energies of commonly used X-ray calibration sources. This requires new calibration sources with sub-keV energies. When searching for nuclear recoil signals, the calibration source should ideally cause mono-energetic nuclear recoils in the relevant energy range. Recently, a new calibration method based on the radiative neutron capture on $^{182}$W with subsequent de-excitation via single $\gamma$-emission leading to a nuclear recoil peak at 112 eV was proposed. The CRESST-III dark matter search operated several CaWO$_{4}$-based detector modules with detection thresholds below 100 eV in the past years. We report the observation of a peak around the expected energy of 112 eV in the data of three different detector modules recorded while irradiated with neutrons from different AmBe calibration sources. We compare the properties of the observed peaks with Geant-4 simulations and assess the prospects of using this for the energy calibration of CRESST-III detectors., Comment: 8 pages, 4 figures; submitted to Phys. Rev. D

We study the complexity of computing equilibria in binary public goods games on undirected graphs. In such a game, players correspond to vertices in a graph and face a binary choice of performing an action, or not. Each player's decision depends only on the number of neighbors in the graph who perform the action and is encoded by a per-player binary pattern. We show that games with decreasing patterns (where players only want to act up to a threshold number of adjacent players doing so) always have a pure Nash equilibrium and that one is reached from any starting profile by following a polynomially bounded sequence of best responses. For non-monotonic patterns of the form $10^k10^*$ (where players want to act alone or alongside $k + 1$ neighbors), we show that it is $\mathsf{NP}$-hard to decide whether a pure Nash equilibrium exists. We further investigate a generalization of the model that permits ties of varying strength: an edge with integral weight $w$ behaves as $w$ parallel edges. While, in this model, a pure Nash equilibrium still exists for decreasing patters, we show that the task of computing one is $\mathsf{PLS}$-complete., Comment: To appear in the Proceedings of the 24th ACM Conference on Economics and Computation (EC 2023)

We present limits on the spin-independent interaction cross section of dark matter particles with silicon nuclei, derived from data taken with a cryogenic calorimeter with 0.35 g target mass operated in the CRESST-III experiment. A baseline nuclear recoil energy resolution of $(1.36\pm 0.05)$ eV$_{\text{nr}}$, currently the lowest reported for macroscopic particle detectors, and a corresponding energy threshold of $(10.0\pm 0.2)$ eV$_{\text{nr}}$ have been achieved, improving the sensitivity to light dark matter particles with masses below 160 MeV/c$^2$ by a factor of up to 20 compared to previous results. We characterize the observed low energy excess, and we exclude noise triggers and radioactive contaminations on the crystal surfaces as dominant contributions., Comment: 8 pages, 5 figures; precised the position of the calibration source in Fig. 1; extended the discussion about the observed energy spectrum; added the DM limit curve to ancillary files. Published in Phys. Rev. D

A challenge in spoken language translation is that plenty of spoken content is long-form, but short units are necessary for obtaining high-quality translations. To address this mismatch, we fine-tune a general-purpose, large language model to split long ASR transcripts into segments that can be independently translated so as to maximize the overall translation quality. We compare to several segmentation strategies and find that our approach improves BLEU score on three languages by an average of 2.7 BLEU overall compared to an automatic punctuation baseline. Further, we demonstrate the effectiveness of two constrained decoding strategies to improve well-formedness of the model output from above 99% to 100%.

We report on novel investigations into training models that make sentences concise. We define the task and show that it is different from related tasks such as summarization and simplification. For evaluation, we release two test sets, consisting of 2000 sentences each, that were annotated by two and five human annotators, respectively. We demonstrate that conciseness is a difficult task for which zero-shot setups with large neural language models often do not perform well. Given the limitations of these approaches, we propose a synthetic data generation method based on round-trip translations. Using this data to either train Transformers from scratch or fine-tune T5 models yields our strongest baselines that can be further improved by fine-tuning on an artificial conciseness dataset that we derived from multi-annotator machine translation test sets., Comment: EMNLP 2022 Workshop on Text Simplification, Accessibility, and Readability (TSAR)

The CRESST experiment employs cryogenic calorimeters for the sensitive measurement of nuclear recoils induced by dark matter particles. The recorded signals need to undergo a careful cleaning process to avoid wrongly reconstructed recoil energies caused by pile-up and read-out artefacts. We frame this process as a time series classification task and propose to automate it with neural networks. With a data set of over one million labeled records from 68 detectors, recorded between 2013 and 2019 by CRESST, we test the capability of four commonly used neural network architectures to learn the data cleaning task. Our best performing model achieves a balanced accuracy of 0.932 on our test set. We show on an exemplary detector that about half of the wrongly predicted events are in fact wrongly labeled events, and a large share of the remaining ones have a context-dependent ground truth. We furthermore evaluate the recall and selectivity of our classifiers with simulated data. The results confirm that the trained classifiers are well suited for the data cleaning task., Comment: 12 pages, 8 figures, 6 tables

CRESST is a leading direct detection sub-$\mathrm{GeVc}^{-2}$ dark matter experiment. During its second phase, cryogenic bolometers were used to detect nuclear recoils off the $\mathrm{CaWO}_4$ target crystal nuclei. The previously established electromagnetic background model relies on secular equilibrium (SE) assumptions. In this work, a validation of SE is attempted by comparing two likelihood-based normalisation results using a recently developed spectral template normalisation method based on Bayesian likelihood. We find deviations from SE; further investigations are necessary to determine their origin., Comment: Part of special issue: ICRM-LLRMT22 8th International Conference on Radionuclide Metrology - Low Level Radioactivity Measurement and Techniques, 6 pages, 3 figures

The CRESST experiment observes an unexplained excess of events at low energies. In the current CRESST-III data-taking campaign we are operating detector modules with different designs to narrow down the possible explanations. In this work, we show first observations of the ongoing measurement, focusing on the comparison of time, energy and temperature dependence of the excess in several detectors. These exclude dark matter, radioactive backgrounds and intrinsic sources related to the crystal bulk as a major contribution., Comment: 10 pages, 5 figures; to be published in IDM2022 proceedings; corrected several minor typos

In the past decades, numerous experiments have emerged to unveil the nature of dark matter, one of the most discussed open questions in modern particle physics. Among them, the CRESST experiment, located at the Laboratori Nazionali del Gran Sasso, operates scintillating crystals as cryogenic phonon detectors. In this work, we present first results from the operation of two detector modules which both have 10.46 g LiAlO$_2$ targets in CRESST-III. The lithium contents in the crystal are $^6$Li, with an odd number of protons and neutrons, and $^7$Li, with an odd number of protons. By considering both isotopes of lithium and $^{27}$Al, we set the currently strongest cross section upper limits on spin-dependent interaction of dark matter with protons and neutrons for the mass region between 0.25 and 1.5 GeV/c$^2$., Comment: 9 pages, 8 figures

Text-editing models have recently become a prominent alternative to seq2seq models for monolingual text-generation tasks such as grammatical error correction, simplification, and style transfer. These tasks share a common trait - they exhibit a large amount of textual overlap between the source and target texts. Text-editing models take advantage of this observation and learn to generate the output by predicting edit operations applied to the source sequence. In contrast, seq2seq models generate outputs word-by-word from scratch thus making them slow at inference time. Text-editing models provide several benefits over seq2seq models including faster inference speed, higher sample efficiency, and better control and interpretability of the outputs. This tutorial provides a comprehensive overview of text-editing models and current state-of-the-art approaches, and analyzes their pros and cons. We discuss challenges related to productionization and how these models can be used to mitigate hallucination and bias, both pressing challenges in the field of text generation., Comment: Accepted as a tutorial at NAACL 2022

We consider the problem of learning generalized policies for classical planning domains using graph neural networks from small instances represented in lifted STRIPS. The problem has been considered before but the proposed neural architectures are complex and the results are often mixed. In this work, we use a simple and general GNN architecture and aim at obtaining crisp experimental results and a deeper understanding: either the policy greedy in the learned value function achieves close to 100% generalization over instances larger than those used in training, or the failure must be understood, and possibly fixed, logically. For this, we exploit the relation established between the expressive power of GNNs and the $C_{2}$ fragment of first-order logic (namely, FOL with 2 variables and counting quantifiers). We find for example that domains with general policies that require more expressive features can be solved with GNNs once the states are extended with suitable "derived atoms" encoding role compositions and transitive closures that do not fit into $C_{2}$. The work follows the GNN approach for learning optimal general policies in a supervised fashion (Stahlberg, Bonet, Geffner, 2022); but the learned policies are no longer required to be optimal (which expands the scope, as many planning domains do not have general optimal policies) and are learned without supervision. Interestingly, value-based reinforcement learning methods that aim to produce optimal policies, do not always yield policies that generalize, as the goals of optimality and generality are in conflict in domains where optimal planning is NP-hard., Comment: Proceedings of the 19th International Conference on Principles of Knowledge Representation and Reasoning (KR-22)

Detection of extremely rare variant alleles, such as tumour DNA, within a complex mixture of DNA molecules is experimentally challenging due to sequencing errors. Barcoding of target DNA molecules in library construction for next-generation sequencing provides a way to identify and bioinformatically remove polymerase induced errors. During the barcoding procedure involving $t$ consecutive PCR cycles, the DNA molecules become barcoded by unique molecular identifiers (UMI). Different library construction protocols utilise different values of $t$. The effect of a larger $t$ and imperfect PCR amplifications is poorly described. This paper proposes a branching process with growing immigration as a model describing the random outcome of $t$ cycles of PCR barcoding. Our model discriminates between five different amplification rates $r_1$, $r_2$, $r_3$, $r_4$, $r$ for different types of molecules associated with the PCR barcoding procedure. We study this model by focussing on $C_t$, the number of clusters of molecules sharing the same UMI, as well as $C_t(m)$, the number of UMI clusters of size $m$. Our main finding is a remarkable asymptotic pattern valid for moderately large $t$. It turns out that $E(C_t(m))/E(C_t)\approx 2^{-m}$ for $m=1,2,\ldots$, regardless of the underlying parameters $(r_1,r_2,r_3,r_4,r)$. The knowledge of the quantities $C_t$ and $C_t(m)$ as functions of the experimental parameters $t$ and $(r_1,r_2,r_3,r_4,r)$ will help the users to draw more adequate conclusions from the outcomes of different sequencing protocols.

The softmax layer in neural machine translation is designed to model the distribution over mutually exclusive tokens. Machine translation, however, is intrinsically uncertain: the same source sentence can have multiple semantically equivalent translations. Therefore, we propose to replace the softmax activation with a multi-label classification layer that can model ambiguity more effectively. We call our loss function Single-label Contrastive Objective for Non-Exclusive Sequences (SCONES). We show that the multi-label output layer can still be trained on single reference training data using the SCONES loss function. SCONES yields consistent BLEU score gains across six translation directions, particularly for medium-resource language pairs and small beam sizes. By using smaller beam sizes we can speed up inference by a factor of 3.9x and still match or improve the BLEU score obtained using softmax. Furthermore, we demonstrate that SCONES can be used to train NMT models that assign the highest probability to adequate translations, thus mitigating the "beam search curse". Additional experiments on synthetic language pairs with varying levels of uncertainty suggest that the improvements from SCONES can be attributed to better handling of ambiguity., Comment: NAACL 2022 paper

In many natural language processing (NLP) tasks the same input (e.g. source sentence) can have multiple possible outputs (e.g. translations). To analyze how this ambiguity (also known as intrinsic uncertainty) shapes the distribution learned by neural sequence models we measure sentence-level uncertainty by computing the degree of overlap between references in multi-reference test sets from two different NLP tasks: machine translation (MT) and grammatical error correction (GEC). At both the sentence- and the task-level, intrinsic uncertainty has major implications for various aspects of search such as the inductive biases in beam search and the complexity of exact search. In particular, we show that well-known pathologies such as a high number of beam search errors, the inadequacy of the mode, and the drop in system performance with large beam sizes apply to tasks with high level of ambiguity such as MT but not to less uncertain tasks such as GEC. Furthermore, we propose a novel exact $n$-best search algorithm for neural sequence models, and show that intrinsic uncertainty affects model uncertainty as the model tends to overly spread out the probability mass for uncertain tasks and sentences., Comment: ACL 2022 paper

Many low-threshold experiments observe sharply rising event rates of yet unknown origins below a few hundred eV, and larger than expected from known backgrounds. Due to the significant impact of this excess on the dark matter or neutrino sensitivity of these experiments, a collective effort has been started to share the knowledge about the individual observations. For this, the EXCESS Workshop was initiated. In its first iteration in June 2021, ten rare event search collaborations contributed to this initiative via talks and discussions. The contributing collaborations were CONNIE, CRESST, DAMIC, EDELWEISS, MINER, NEWS-G, NUCLEUS, RICOCHET, SENSEI and SuperCDMS. They presented data about their observed energy spectra and known backgrounds together with details about the respective measurements. In this paper, we summarize the presented information and give a comprehensive overview of the similarities and differences between the distinct measurements. The provided data is furthermore publicly available on the workshop's data repository together with a plotting tool for visualization., Comment: 44 pages, 20 figures; Editors: A. Fuss, M. Kaznacheeva, F. Reindl, F. Wagner; updated copyright statements and funding information

Text normalization, or the process of transforming text into a consistent, canonical form, is crucial for speech applications such as text-to-speech synthesis (TTS). In TTS, the system must decide whether to verbalize "1995" as "nineteen ninety five" in "born in 1995" or as "one thousand nine hundred ninety five" in "page 1995". We present an experimental comparison of various Transformer-based sequence-to-sequence (seq2seq) models of text normalization for speech and evaluate them on a variety of datasets of written text aligned to its normalized spoken form. These models include variants of the 2-stage RNN-based tagging/seq2seq architecture introduced by Zhang et al. (2019), where we replace the RNN with a Transformer in one or more stages, as well as vanilla Transformers that output string representations of edit sequences. Of our approaches, using Transformers for sentence context encoding within the 2-stage model proved most effective, with the fine-tuned BERT encoder yielding the best performance.

CRESST is one of the most prominent direct detection experiments for dark matter particles with sub-GeV/c$^2$ mass. One of the advantages of the CRESST experiment is the possibility to include a large variety of nuclides in the target material used to probe dark matter interactions. In this work, we discuss in particular the interactions of dark matter particles with protons and neutrons of $^{6}$Li. This is now possible thanks to new calculations on nuclear matrix elements of this specific isotope of Li. To show the potential of using this particular nuclide for probing dark matter interactions, we used the data collected previously by a CRESST prototype based on LiAlO$_2$ and operated in an above ground test-facility at Max-Planck-Institut f\"ur Physik in Munich, Germany. In particular, the inclusion of $^{6}$Li in the limit calculation drastically improves the result obtained for spin-dependent interactions with neutrons in the whole mass range. The improvement is significant, greater than two order of magnitude for dark matter masses below 1 GeV/c$^2$, compared to the limit previously published with the same data.