Search

A predominantly electric E\&m storage ring, with weak superimposed magnetic bending, is shown to be capable of storing two different particle type bunches, such as helion (h) and deuteron (d), or h and electron ($e^-$), co-traveling with different velocities on the same central orbit. Rear-end collisions occurring periodically in a full acceptance particle detector/polarimeter, allow the (previously inaccessible) direct measurement of the spin dependence of nuclear transmutation for center of mass (CM) kinetic energies (KE) ranging from hundreds of keV up toward pion production thresholds. With the nuclear process occurring in a semi-relativistic moving frame, all initial and final state particles have convenient laboratory frame KEs in the tens to hundreds of MeV. The rear-end collisions occur as faster stored bunches pass through slower bunches. An inexpensive facility capable of meeting these requirements is described, with several nuclear channels as examples. Especially noteworthy are the $e^{+/-}$-induced weak interaction triton (t) $\beta$-decay processes, t + $e^+ \rightarrow$ h + $\nu$ and h + $e^- \rightarrow$ t + $\nu$. Experimental capability of measurement of the spin dependence of the induced triton case is emphasized. For cosmological nuclear physics, the experimental improvement will be produced by the storage ring's capability to investigate the spin dependence of nuclear transmutation processes at reduced kinetic energies compared to what can be obtained with fixed target geometry., Comment: Intended for submission to JINST subsequent to HB 2023 Workshop

The pretraining of state-of-the-art large language models now requires trillions of words of text, which is orders of magnitude more than available for the vast majority of languages. While including text in more than one language is an obvious way to acquire more pretraining data, multilinguality is often seen as a curse, and most model training efforts continue to focus near-exclusively on individual large languages. We believe that multilinguality can be a blessing and that it should be possible to substantially improve over the capabilities of monolingual models for small languages through multilingual training. In this study, we introduce Poro 34B, a 34 billion parameter model trained for 1 trillion tokens of Finnish, English, and programming languages, and demonstrate that a multilingual training approach can produce a model that not only substantially advances over the capabilities of existing models for Finnish, but also excels in translation and is competitive in its class in generating English and programming languages. We release the model parameters, scripts, and data under open licenses at https://huggingface.co/LumiOpen/Poro-34B.

This paper resurrects and archives an unpublished original Cornell Laboratory of Nuclear Studies report by Yuko Okamoto and Richard Talman, "Rational Approximation of the Complex Error Function and the Electric Field of a Two-Dimensional Gaussian Charge Distribution" CBN 80-13, dating from September 1980, during the start-up period of the CESR-CLEO $e^+e^-$ collider. This code has played a significant role in the calculation of the beam-beam interaction (in particular the beam-beam tune shift) for subsequent storage ring colliders. Electronic access to the (refactored) original codes is provided by active links., Comment: 43 pages, 2 figures

This paper introduces Bayesian uncertainty modeling using Stochastic Weight Averaging-Gaussian (SWAG) in Natural Language Understanding (NLU) tasks. We apply the approach to standard tasks in natural language inference (NLI) and demonstrate the effectiveness of the method in terms of prediction accuracy and correlation with human annotation disagreements. We argue that the uncertainty representations in SWAG better reflect subjective interpretation and the natural variation that is also present in human language understanding. The results reveal the importance of uncertainty modeling, an often neglected aspect of neural language modeling, in NLU tasks., Comment: NoDaLiDa 2023 camera ready

Of all nuclear physics experiments none are more fundamental than ``elastic'' $p,p$ and, secondarily, $p,d$ or $d,d$ scattering. Recognizing that these particles are themselves composite, ``elastic'' scattering may be accompanied by temporary internal rearrangement with undetectably small energy loss. % This paper argues that correct calculation of the spin dependence of $p,p$ elastic scattering must account for a previously-neglected relativistic effect of ``$G$'', the anomalous magnetic dipole moment (MDM) of the scattering particles and proceeds to describe storage ring $p,p$ scattering configurations capable of confirming this contention. % Especially important experimentally for protons is the existence of ``near perfect'' proton-carbon scattering polarimetric analyzing power $A$ near $K=183.1$\,MeV laboratory kinetic energy. % Possibilities: (i) With counter-circulating $K=200\,$MeV energy proton beams the final spin states of coincident scattered protons can be determined with high probability for a significantly large fraction of all scatters. For comparison with current descriptions this corresponds roughly to proton kinetic energy $K=400\,$MeV in the laboratory frame, close to the pion production threshold. % (ii) In a ``DERBENEV-style'' figure-8 storage ring, independently polarized, diametrically opposite bunches on orthogonal orbits can collide at the beam crossover point with symmetric $K''\approx200$ MeV energies in a slow, transversely moving frame. % (iii) $p$ and $d$ beams can counter-circulate at the same time in a small racetrack shaped ring with superimposed electric and magnetic bending. In this case the scattering would be ``WOLFENSTEIN-style'' with collinear incident orbits. % To investigate the consistency of quantum mechanics and special relativity it is proposed to implement options (ii) and (iii) in the COSY beam hall., Comment: PSTP2022 Workshop, Mainz, Germany

The upper limit on (time reversal symmetry T-violating) permanent hadron electric dipole moments (EDMs) is the PSI neutron EDM value; $d_n = (0.0\pm1.1_{\rm stat}\pm0.2_{\rm sys}\times10^{-26})\,e$\,cm. This paper describes an experiment to be performed at a BNL-proposed CLIP project which is to be capable of producing intense polarized beams of protons, $p$, helions (He${}^3$ nuclei), h, and other isotopes. The EDM prototype ring PTR (proposed at COSY Lab, Juelich) is expected to measure individual particle EDMs (for example ${\rm EDM\_p}$ for the proton) using simultaneous counter-rotating polarized proton beams, with statistical error $\pm10^{-30}$e.cm after one year running time, four orders of magnitude less than the PSI neutron EDM upper limit, and with comparable systematic error. A composite particle, the helion faces T-symmetry constraints more challenging than the proton. Any measurably large value of $$\Delta={\rm EDM}_h-{\rm EDM}_p,$$ the difference of helion and proton EDMs, would represent BSM physics. The plan is to replicate PTR at BNL. The dominant systematic error would be canceled two ways, both made possible by phase-locking "doubly-magic" 38.6\,MeV proton and 39.2\,MeV helion spin tunes. This stabilizes their MDM-induced in-plane precessions, without affecting their EDM-induced out-of-plane precessions. The dominant systematic error would therefore cancel in the meaurement of $\Delta$ in a fixed field configuration. Another systematic error cancellation will come from averaging runs for which both magnetic field and beam circulation directions are reversed. Precise magnetic field reversal is made possible by the reproducible absolute frequency phase-locking over long runs to eliminate the need for (impractically precise) magnetic field measurement. Risk of EDM measurement failure is discussed in a final appendix.

A central question in natural language understanding (NLU) research is whether high performance demonstrates the models' strong reasoning capabilities. We present an extensive series of controlled experiments where pre-trained language models are exposed to data that have undergone specific corruption transformations. These involve removing instances of specific word classes and often lead to non-sensical sentences. Our results show that performance remains high on most GLUE tasks when the models are fine-tuned or tested on corrupted data, suggesting that they leverage other cues for prediction even in non-sensical contexts. Our proposed data transformations can be used to assess the extent to which a specific dataset constitutes a proper testbed for evaluating models' language understanding capabilities., Comment: *SEM 2022 camera ready version

In contrast to a "single particle table-top trap", an essential feature of a storage ring "trap" is that $10^{10}$ or more particles can have their spins aligned in a polarized beam. This is a nunber of polarized particles large enough for the beam polarization to be detected externally, and fed back to permit external control of the beam polarization. The level of achievable spin control, though classical, not quantum mechanical, can be comparable to the control of one or a small number of polarized particles in a low energy trap. Motivated to investigate time reversal invariance, especially the detection of non-zero electric dipole moments (EDMs) this paper describes the design of a low energy storage ring having the superimposed electric and magnetic bending needed to "freeze" the spins of polarized beams. For electrons (of either sign) and protons the spins can be frozen with all-electric bending but, in general, superimposed electric/magnetic bending is required. Since constructive bending superposition in one direction implies destructive superposition in the other direction, counter-circulating beams must differ, either in particle type or momentum, in order for their orbits to be identical. For globally frozen spin operation the bunch polarizations remain constant relative to the momenta, for example remaining parallel to the circuating beam momentum vectors. With superimposed electric and magnetic bending, the globally frozen spin condition can be met over a continua (specific to particle type) of E/B ratios. When this condition is met, the out-of-plane, EDM-induced precession accumulates monitonically, which is obligatory for producing a measurably large EDM signal. As Koop has explained, the EDM signal will still accumulate if the polarization is allowed to "roll like a wheel" around a radial axis., Comment: Accepted for publication, 24 August 2021, in ICFA Beam Dynamics Newsletter #82, Advanced Accelerator Modelling

Pre-trained neural language models give high performance on natural language inference (NLI) tasks. But whether they actually understand the meaning of the processed sequences remains unclear. We propose a new diagnostics test suite which allows to assess whether a dataset constitutes a good testbed for evaluating the models' meaning understanding capabilities. We specifically apply controlled corruption transformations to widely used benchmarks (MNLI and ANLI), which involve removing entire word classes and often lead to non-sensical sentence pairs. If model accuracy on the corrupted data remains high, then the dataset is likely to contain statistical biases and artefacts that guide prediction. Inversely, a large decrease in model accuracy indicates that the original dataset provides a proper challenge to the models' reasoning capabilities. Hence, our proposed controls can serve as a crash test for developing high quality data for NLI tasks., Comment: NoDaLiDa 2021 camera ready

The J\"ulich Electric Dipole moment Investigation (JEDI) collaboration aims at a direct measurement of the Electric Dipole Moment (EDM) of protons and deuterons using a storage ring. The measurement is based on a polarization measurement. In order to reach highest accuracy, one has to know the exact trajectory through the magnets, especially the quadrupoles, to avoid the influence of magnetic fields on the polarization vector. In this paper, the development of a beam-based alignment technique is described that was developed and implemented at the COoler SYnchrotron (COSY) at Forschungszentrum J\"ulich. Well aligned quadrupoles permit one to absolutely calibrate the Beam Position Monitors (BPMs). The method is based on the fact that a particle beam, which does not pass through the center of a quadrupole, experiences a deflection. The precision reached by the method is approximately 40 micro meter. Some consequences for the design of a new high precision storage ring for EDM mesasurements are discussed., Comment: Submitted to "Journal of Instrumentation (JINST)"

A measurement of vector analyzing powers in elastic deuteron-carbon scattering has been performed at the Cooler Synchrotron COSY of Forschungszentrum J\"ulich, Germany. Seven kinetic beam energies between 170 and 380 MeV have been used. A vector-polarized beam from a polarized deuteron source was injected, accelerated to the final desired energy and stored in COSY. A thin needle-shaped diamond strip was used as a carbon target, onto which the beam was slowly steered. Elastically scattered deuterons were identified in the forward direction using various layers of scintillators and straw tubes. Where data exist in the literature (at 200 and 270 MeV), excellent agreement of the angular shape was found. The beam polarization of the presented data was deduced by fitting the absolute scale of the analyzing power to these references. Our results extend the world data set and are necessary for polarimetry of future electric dipole moment searches at storage rings. They will as well serve as an input for theoretical description of polarized hadron-hadron scattering.

A previous publication has shown that the in-plane polarization (IPP) component of a polarized 0.97-GeV/c deuteron beam in the COSY storage ring may acquire a polarization half-life in excess of 1000 s through a combination of beam bunching, electron cooling (prior to any spin manipulation), sextupole field adjustment, and a limitation of the beam intensity. This paper documents further tests pointing to additional gains in the IPP lifetime if cooling is active throughout the beam store., Comment: 11 pages, 6 figures Submitted to Physical Review Accelerators and Beams

The proposed method exploits charged particles confined as a storage ring beam (proton, deuteron, possibly $^3$He) to search for an intrinsic electric dipole moment (EDM) aligned along the particle spin axis. Statistical sensitivities could approach 10$^{-29}$ e$\cdot$cm. The challenge will be to reduce systematic errors to similar levels. The ring will be adjusted to preserve the spin polarisation, initially parallel to the particle velocity, for times in excess of 15 minutes. Large radial electric fields, acting through the EDM, will rotate the polarisation from the longitudinal to the vertical direction. The slow rise in the vertical polarisation component, detected through scattering from a target, signals the EDM. The project strategy is outlined. A stepwise plan is foreseen, starting with ongoing COSY activities that demonstrate technical feasibility. Achievements to date include reduced polarization measurement errors, long horizontal plane polarization lifetimes, and control of the polarization direction through feedback from scattering measurements. The project continues with a proof-of-capability measurement (precursor experiment; first direct deuteron EDM measurement), an intermediate prototype ring (proof-of-principle; demonstrator for key technologies), and finally a high-precision electric-field storage ring., Comment: 243 pages

In this paper we introduce a new natural language processing dataset and benchmark for predicting prosodic prominence from written text. To our knowledge this will be the largest publicly available dataset with prosodic labels. We describe the dataset construction and the resulting benchmark dataset in detail and train a number of different models ranging from feature-based classifiers to neural network systems for the prediction of discretized prosodic prominence. We show that pre-trained contextualized word representations from BERT outperform the other models even with less than 10% of the training data. Finally we discuss the dataset in light of the results and point to future research and plans for further improving both the dataset and methods of predicting prosodic prominence from text. The dataset and the code for the models are publicly available., Comment: NoDaLiDa 2019 camera ready

In this paper, we present the University of Helsinki submissions to the WMT 2019 shared task on news translation in three language pairs: English-German, English-Finnish and Finnish-English. This year, we focused first on cleaning and filtering the training data using multiple data-filtering approaches, resulting in much smaller and cleaner training sets. For English-German, we trained both sentence-level transformer models and compared different document-level translation approaches. For Finnish-English and English-Finnish we focused on different segmentation approaches, and we also included a rule-based system for English-Finnish., Comment: To appear in WMT19

This project exploits charged particles confined as a storage ring beam (proton, deuteron, possibly $^3$He) to search for an intrinsic electric dipole moment (EDM, $\vec d$) aligned along the particle spin axis. Statistical sensitivities can approach $10^{-29}$~e$\cdot$cm. The challenge will be to reduce systematic errors to similar levels. The ring will be adjusted to preserve the spin polarization, initially parallel to the particle velocity, for times in excess of 15 minutes. Large radial electric fields, acting through the EDM, will rotate the polarization ($\vec d \times\vec E$). The slow rise in the vertical polarization component, detected through scattering from a target, signals the EDM. The project strategy is outlined. It foresees a step-wise plan, starting with ongoing COSY activities that demonstrate technical feasibility. Achievements to date include reduced polarization measurement errors, long horizontal-plane polarization lifetimes, and control of the polarization direction through feedback from the scattering measurements. The project continues with a proof-of-capability measurement (precursor experiment; first direct deuteron EDM measurement), an intermediate prototype ring (proof-of-principle; demonstrator for key technologies), and finally the high precision electric-field storage ring.

This paper discusses "doubly-magic trap" operation of storage rings with superimposed electric and magnetic bending, to allow the spins of beams of different particle type to be frozen (i.e. both magic) simultaneously. A few doubly-magic cases have been found; one has 86.62990502 MeV frozen spin protons and 30.09255159 MeV frozen spin positrons (with accuracies matching their known magnetic moments) counter-circulating in the same storage ring. (Assuming zero positron EDM value) the positron beam can be used to null the radial magnetic field, enabling the proton EDM to be measured free of the dominant systematic error. The same trick, using 38.6 MeV protons and 39.2 MeV helions (He3) yields the difference of proton and helion EDMs---any measureably large difference would represent physics beyond the standard model---using an inexpensive 10 m bending radius storage ring, and already-proven operational spin control technology.

Neural network models have been very successful in natural language inference, with the best models reaching 90% accuracy in some benchmarks. However, the success of these models turns out to be largely benchmark specific. We show that models trained on a natural language inference dataset drawn from one benchmark fail to perform well in others, even if the notion of inference assumed in these benchmarks is the same or similar. We train six high performing neural network models on different datasets and show that each one of these has problems of generalizing when we replace the original test set with a test set taken from another corpus designed for the same task. In light of these results, we argue that most of the current neural network models are not able to generalize well in the task of natural language inference. We find that using large pre-trained language models helps with transfer learning when the datasets are similar enough. Our results also highlight that the current NLI datasets do not cover the different nuances of inference extensively enough., Comment: Accepted to the 2019 ACL Workshop BackboxNLP: Analyzing and interpreting neural networks for NLP

Sentence-level representations are necessary for various NLP tasks. Recurrent neural networks have proven to be very effective in learning distributed representations and can be trained efficiently on natural language inference tasks. We build on top of one such model and propose a hierarchy of BiLSTM and max pooling layers that implements an iterative refinement strategy and yields state of the art results on the SciTail dataset as well as strong results for SNLI and MultiNLI. We can show that the sentence embeddings learned in this way can be utilized in a wide variety of transfer learning tasks, outperforming InferSent on 7 out of 10 and SkipThought on 8 out of 9 SentEval sentence embedding evaluation tasks. Furthermore, our model beats the InferSent model in 8 out of 10 recently published SentEval probing tasks designed to evaluate sentence embeddings' ability to capture some of the important linguistic properties of sentences., Comment: To appear in JNLE