Search

Large language models (LMs) are capable of in-context learning from a few demonstrations (example-label pairs) to solve new tasks during inference. Despite the intuitive importance of high-quality demonstrations, previous work has observed that, in some settings, ICL performance is minimally affected by irrelevant labels (Min et al., 2022). We hypothesize that LMs perform ICL with irrelevant labels via two sequential processes: an inference function that solves the task, followed by a verbalization function that maps the inferred answer to the label space. Importantly, we hypothesize that the inference function is invariant to remappings of the label space (e.g., "true"/"false" to "cat"/"dog"), enabling LMs to share the same inference function across settings with different label words. We empirically validate this hypothesis with controlled layer-wise interchange intervention experiments. Our findings confirm the hypotheses on multiple datasets and tasks (natural language inference, sentiment analysis, and topic classification) and further suggest that the two functions can be localized in specific layers across various open-sourced models, including GEMMA-7B, MISTRAL-7B-V0.3, GEMMA-2-27B, and LLAMA-3.1-70B., Comment: EMNLP 2024 Findings

Instruction tuning commonly means finetuning a language model on instruction-response pairs. We discover two forms of adaptation (tuning) that are deficient compared to instruction tuning, yet still yield instruction following; we call this implicit instruction tuning. We first find that instruction-response pairs are not necessary: training solely on responses, without any corresponding instructions, yields instruction following. This suggests pretrained models have an instruction-response mapping which is revealed by teaching the model the desired distribution of responses. However, we then find it's not necessary to teach the desired distribution of responses: instruction-response training on narrow-domain data like poetry still leads to broad instruction-following behavior like recipe generation. In particular, when instructions are very different from those in the narrow finetuning domain, models' responses do not adhere to the style of the finetuning domain. To begin to explain implicit instruction tuning, we hypothesize that very simple changes to a language model's distribution yield instruction following. We support this by hand-writing a rule-based language model which yields instruction following in a product-of-experts with a pretrained model. The rules are to slowly increase the probability of ending the sequence, penalize repetition, and uniformly change 15 words' probabilities. In summary, adaptations made without being designed to yield instruction following can do so implicitly.

Low-mass X-ray binaries (LMXBs) with neutron stars show quite different features which depend on the rate of mass transfer from the donor star. With a high transfer rate the Z sources are in a persistent soft spectral state, with a moderate rate the transient Atoll sources have outburst cycles like the black hole X-ray binaries. The observations document very long outburst recurrence times for quite a number of sources. We follow with our computations the evolution of the accretion disc until the onset of the ionization instability. For sources with a low mass transfer rate the accumulation of matter in the disc is essentially reduced due to the continuous evaporation of matter from the disc to the coronal flow. Different mass transfer rates result in nearly the same amount of matter accumulated for the outburst which means the outburst properties are similar for sources with short and sources with long outburst cycles, contrary to some expectations. Then of systems with long recurrence time less sources will be detected and the total population of LMXBs could be larger than it appears. This would relieve the apparent problem that the observed number of LMXBs as progenitors of millisecond pulsars (MSP) is too small compared to the number of MSP. Concerning the few quasi-persistent sources with year-long soft states we argue that these states are not outbursts, but quasi-stationary hot states as in Z sources., Comment: 7 pages, 4 figures, published in MNRAS

Among the 3d transition metal ions doped polar Ni3TeO6, Mn-doped Ni3TeO6 has stimulated great interest due to its high magnetic ordering temperature and complex magnetic phases, but the mechanism of magnetoelectric (ME) coupling is far from understood. Herein we report our systematic investigation of the chemical control of magnetism, metamagnetic transition, and ME properties of Ni3-xMnxTeO6 single crystals in high magnetic field (H) up to 52 T. We present a previously unreported weak ferromagnetic behavior appeared in the ab plane below 9.5 K in addition to the incommensurate helical and commensurate collinear antiferromagnetic states. In the low-field region, a spin-flop type metamagnetic transition without any hysteresis occurs at Hc1 for H // c, while another metamagnetic transition accompanied with a change in electric polarization is observed at Hc2 in the high-field region both for H // c and H // ab above 30 K, which can be attributed to the sudden rotation of magnetic moments at Ni2 sites. The ME measurements reveal that a first-order ME effect is observed in the low-T and low-H regions, while a second-order ME coupling term appears above 30 K in the magnetic field range of Hc1 < H < Hc2 for H // c and H < Hc2 for H // ab, both becoming significant with increasing temperature. Eventually, they are dominated by the second-order ME effect near the antiferromagnetic transition temperature. The present work demonstrates that Ni3-xMnxTeO6 is an exotic magnetoelectric material compared with Ni3TeO6 and its derivatives, thereby providing insights to better understand the magnetism and ME coupling in Ni3TeO6 and its derivatives., Comment: 30 pages with 8 figures

The van der Waals (vdW) layered multiferroics, which host simultaneous ferroelectric and magnetic orders, have attracted attention not only for their potentials to be utilized in nanoelectric devices and spintronics, but also offer alternative opportunities for emergent physical phenomena. To date, the vdW layered multiferroic materials are still very rare. In this work, we have investigated the magnetic structure and magnetoelectric effects in Co5(TeO3)4Cl2, a promising new multiferroic compound with antiferromagnetic (AFM) Neel point TN = 18 K. The neutron powder diffraction reveals the non-coplanar AFM state with preferred Neel vector along the c-axis, while a spin re-orientation occurring between 8 K and 15 K is identified, which results from the distinct temperature dependence of the non-equivalent Co sites moment in Co5(TeO3)4Cl2. What is more, it is found that Co5(TeO3)4Cl2 is one of the best vdW multiferroics studied so far in terms of the multiferroic performance. The measured linear ME coefficient exhibits the emergent oscillation dependence of the angle between magnetic field and electric field, and the maximal value is as big as 45 ps/m. It is suggested that Co5(TeO3)4Cl2 is an appreciated platform for exploring the emergent multiferroicity in vdW layered compounds., Comment: 31 pages, 9 figures

Quantum machine learning (QML) is one of the most promising applications of quantum computation. However, it is still unclear whether quantum advantages exist when the data is of a classical nature and the search for practical, real-world applications of QML remains active. In this work, we apply the well-studied quantum support vector machine (QSVM), a powerful QML model, to a binary classification task which classifies peptides as either hemolytic or non-hemolytic. Using three peptide datasets, we apply and contrast the performance of the QSVM, numerous classical SVMs, and the best published results on the same peptide classification task, out of which the QSVM performs best. The contributions of this work include (i) the first application of the QSVM to this specific peptide classification task, (ii) an explicit demonstration of QSVMs outperforming the best published results attained with classical machine learning models on this classification task and (iii) empirical results showing that the QSVM is capable of outperforming many (and possibly all) classical SVMs on this classification task. This foundational work paves the way to verifiable quantum advantages in the field of computational biology and facilitates safer therapeutic development., Comment: 15 pages, 4 figures

Large-scale structure (LSS) surveys will increasingly provide stringent constraints on our cosmological models. Recently, the density-marked correlation function (MCF) has been introduced, offering an easily computable density-correlation statistic. Simulations have demonstrated that MCFs offer additional, independent constraints on cosmological models beyond the standard two-point correlation (2PCF). In this study, we apply MCFs for the first time to SDSS CMASS data, aiming to investigate the statistical information regarding clustering and anisotropy properties in the Universe and assess the performance of various weighting schemes in MCFs. Upon analyzing the CMASS data, we observe that, by combining different weights ($\alpha = [-0.2, 0, 0.2, 0.6]$), the MCFs provide a tight and independent constraint on the cosmological parameter $\Omega_m$, yielding $\Omega_m = 0.293 \pm0.006$ at the $1\sigma$ level, which represents a significant reduction in the statistical error by a factor of 3.4 compared to that from 2PCF. Our constraint is consistent with recent findings from the small-scale clustering of BOSS galaxies \cite{arXiv:2203.08999v2} within the 1$\sigma$ level. However, we also find that our estimate is lower than the Planck measurements by about 2.6$\sigma$, indicating the potential presence of new physics beyond the standard cosmological model if all the systematics are fully corrected. The method outlined in this study can be extended to other surveys and datasets, allowing for the constraint of other cosmological parameters. Additionally, it serves as a valuable tool for forthcoming emulator analysis on the Chinese Space Station Telescope (CSST)., Comment: 16 pages, 10 figures

It is widely accepted that X-ray emission in luminous active galactic nuclei (AGNs) originates from hot corona. To prevent the corona from over-cooling by strong X-ray emission, steady heating to the corona is essential, for which the most promising mechanisms is the magnetic reconnection. Detailed studies of the coupled disc and corona, in the frame of magnetic field transferring accretion-released energy from the disc to the corona, reveal that the thermal electrons can only produce X-ray spectrum with $\Gamma_{\rm 2-10\,keV}>2.1$, which is an inevitable consequence of the radiative coupling of the thermal corona and disc. In the present work, we develop the magnetic-reconnection-heated corona model by taking into account the potential non-thermal electrons accelerated in the magnetic reconnection process, in addition to the thermal electrons. We show that the features of the structure and spectrum of the coupled disc and corona can be affected by the fraction of magnetic energy allocated to thermal electrons. Furthermore, we investigate the effects of the power-law index and energy range of non-thermal electrons and the magnetic field on the spectrum. It is found that the X-ray spectrum from the Comptonization of the hybrid electrons can be flatter than that from thermal electrons only, in agreement with observations. By comparing with the observed hard X-ray data, we suggest that a large fraction ($>40\%$) of the magnetic energy be allocated to the non-thermal electrons in the luminous and flat X-ray spectrum AGNs., Comment: 11 pages, 8 figures; Accepted for publication in MNRAS

Fundamental neutron physics, combining precision measurements and theory, probes particle physics at short range with reach well beyond the highest energies probed by the LHC. Significant US efforts are underway that will probe BSM CP violation with orders of magnitude more sensitivity, provide new data on the Cabibbo anomaly, more precisely measure the neutron lifetime and decay, and explore hadronic parity violation. World-leading results from the US Fundamental Neutron Physics community since the last Long Range Plan, include the world's most precise measurement of the neutron lifetime from UCN$\tau$, the final results on the beta-asymmetry from UCNA and new results on hadronic parity violation from the NPDGamma and n-${^3}$He runs at the FNPB (Fundamental Neutron Physics Beamline), precision measurement of the radiative neutron decay mode and n-${}^4$He at NIST. US leadership and discovery potential are ensured by the development of new high-impact experiments including BL3, Nab, LANL nEDM and nEDM@SNS. On the theory side, the last few years have seen results for the neutron EDM from the QCD $\theta$ term, a factor of two reduction in the uncertainty for inner radiative corrections in beta-decay which impacts CKM unitarity, and progress on {\it ab initio} calculations of nuclear structure for medium-mass and heavy nuclei which can eventually improve the connection between nuclear and nucleon EDMs. In order to maintain this exciting program and capitalize on past investments while also pursuing new ideas and building US leadership in new areas, the Fundamental Neutron Physics community has identified a number of priorities and opportunities for our sub-field covering the time-frame of the last Long Range Plan (LRP) under development. This white paper elaborates on these priorities., Comment: arXiv admin note: text overlap with arXiv:2304.03451

While recent language models have the ability to take long contexts as input, relatively little is known about how well they use longer context. We analyze the performance of language models on two tasks that require identifying relevant information in their input contexts: multi-document question answering and key-value retrieval. We find that performance can degrade significantly when changing the position of relevant information, indicating that current language models do not robustly make use of information in long input contexts. In particular, we observe that performance is often highest when relevant information occurs at the beginning or end of the input context, and significantly degrades when models must access relevant information in the middle of long contexts, even for explicitly long-context models. Our analysis provides a better understanding of how language models use their input context and provides new evaluation protocols for future long-context language models., Comment: 18 pages, 16 figures. Accepted for publication in Transactions of the Association for Computational Linguistics (TACL), 2023

Diffusion models (DMs) are one of the most widely used generative models for producing high quality images. However, a flurry of recent papers points out that DMs are least private forms of image generators, by extracting a significant number of near-identical replicas of training images from DMs. Existing privacy-enhancing techniques for DMs, unfortunately, do not provide a good privacy-utility tradeoff. In this paper, we aim to improve the current state of DMs with differential privacy (DP) by adopting the \textit{Latent} Diffusion Models (LDMs). LDMs are equipped with powerful pre-trained autoencoders that map the high-dimensional pixels into lower-dimensional latent representations, in which DMs are trained, yielding a more efficient and fast training of DMs. Rather than fine-tuning the entire LDMs, we fine-tune only the $\textit{attention}$ modules of LDMs with DP-SGD, reducing the number of trainable parameters by roughly $90\%$ and achieving a better privacy-accuracy trade-off. Our approach allows us to generate realistic, high-dimensional images (256x256) conditioned on text prompts with DP guarantees, which, to the best of our knowledge, has not been attempted before. Our approach provides a promising direction for training more powerful, yet training-efficient differentially private DMs, producing high-quality DP images. Our code is available at https://anonymous.4open.science/r/DP-LDM-4525.

Internet links enable users to deepen their understanding of a topic by providing convenient access to related information. However, the majority of links are unanchored -- they link to a target webpage as a whole, and readers may expend considerable effort localizing the specific parts of the target webpage that enrich their understanding of the link's source context. To help readers effectively find information in linked webpages, we introduce the task of anchor prediction, where the goal is to identify the specific part of the linked target webpage that is most related to the source linking context. We release the AuthorAnchors dataset, a collection of 34K naturally-occurring anchored links, which reflect relevance judgments by the authors of the source article. To model reader relevance judgments, we annotate and release ReaderAnchors, an evaluation set of anchors that readers find useful. Our analysis shows that effective anchor prediction often requires jointly reasoning over lengthy source and target webpages to determine their implicit relations and identify parts of the target webpage that are related but not redundant. We benchmark a performant T5-based ranking approach to establish baseline performance on the task, finding ample room for improvement., Comment: 10 pages, 2 figures

Generative search engines directly generate responses to user queries, along with in-line citations. A prerequisite trait of a trustworthy generative search engine is verifiability, i.e., systems should cite comprehensively (high citation recall; all statements are fully supported by citations) and accurately (high citation precision; every cite supports its associated statement). We conduct human evaluation to audit four popular generative search engines -- Bing Chat, NeevaAI, perplexity.ai, and YouChat -- across a diverse set of queries from a variety of sources (e.g., historical Google user queries, dynamically-collected open-ended questions on Reddit, etc.). We find that responses from existing generative search engines are fluent and appear informative, but frequently contain unsupported statements and inaccurate citations: on average, a mere 51.5% of generated sentences are fully supported by citations and only 74.5% of citations support their associated sentence. We believe that these results are concerningly low for systems that may serve as a primary tool for information-seeking users, especially given their facade of trustworthiness. We hope that our results further motivate the development of trustworthy generative search engines and help researchers and users better understand the shortcomings of existing commercial systems., Comment: 25 pages, 12 figures; to appear in Findings of EMNLP 2023

This whitepaper presents the research priorities decided on by attendees of the 2022 Town Meeting for Fundamental Symmetries, Neutrons and Neutrinos, which took place December 13-15, 2022 in Chapel Hill, NC, as part of the Nuclear Science Advisory Committee (NSAC) 2023 Long Range Planning process. A total of 275 scientists registered for the meeting. The whitepaper makes a number of explicit recommendations and justifies them in detail.

This White Paper presents the community inputs and scientific conclusions from the Hot and Cold QCD Town Meeting that took place September 23-25, 2022 at MIT, as part of the Nuclear Science Advisory Committee (NSAC) 2023 Long Range Planning process. A total of 424 physicists registered for the meeting. The meeting highlighted progress in Quantum Chromodynamics (QCD) nuclear physics since the 2015 LRP (LRP15) and identified key questions and plausible paths to obtaining answers to those questions, defining priorities for our research over the coming decade. In defining the priority of outstanding physics opportunities for the future, both prospects for the short (~ 5 years) and longer term (5-10 years and beyond) are identified together with the facilities, personnel and other resources needed to maximize the discovery potential and maintain United States leadership in QCD physics worldwide. This White Paper is organized as follows: In the Executive Summary, we detail the Recommendations and Initiatives that were presented and discussed at the Town Meeting, and their supporting rationales. Section 2 highlights major progress and accomplishments of the past seven years. It is followed, in Section 3, by an overview of the physics opportunities for the immediate future, and in relation with the next QCD frontier: the EIC. Section 4 provides an overview of the physics motivations and goals associated with the EIC. Section 5 is devoted to the workforce development and support of diversity, equity and inclusion. This is followed by a dedicated section on computing in Section 6. Section 7 describes the national need for nuclear data science and the relevance to QCD research., Comment: QCD Town Meeting White Paper, as submitted to 2023 NSAC LRP committee on Feb. 28, 2023

TOI-1695 is a V-mag=13 M-dwarf star from the northern hemisphere at 45$\,$pc from the Sun, around which a 3.134-day periodic transit signal from a super-Earth candidate was identified in TESS photometry. With a transit depth of 1.3$\,$mmag, the radius of candidate TOI-1695.01 was estimated by the TESS pipeline to be 1.82$\,$R$_\oplus$ with an equilibrium temperature of $\sim 620\,$K. We successfully detect a reflex motion of the star and establish it is due to a planetary companion at an orbital period consistent with the photometric transit period thanks to a year-long radial-velocity monitoring of TOI-1695 by the SPIRou infrared spectropolarimeter. We use and compare different methods to reduce and analyse those data. We report a 5.5-$\sigma$ detection of the planetary signal, giving a mass of $5.5 \pm 1.0\,$M$_\oplus$ and a radius of $2.03 \pm 0.18\,$R$_\oplus$. We derive a mean equilibrium planet temperature of $590 \pm 90\,$K. The mean density of this small planet of $3.6 \pm 1.1\,$g$\,$cm$^{-3}$ is similar (1.7-$\sigma$ lower) than that of the Earth. It leads to a non-negligible fraction of volatiles in its atmosphere with $f_{H,He}=0.28^{+0.46}_{-0.23}$% or $f_\text{water}=23 \pm 12$%. TOI-1695$\,$b is a new sub-Neptune planet at the border of the M-dwarf radius valley that can help test formation scenarios for super-Earth/sub-Neptune-like planets., Comment: 15 pages (+9 appendix pages), 19 figures, 8 tables, accepted for publication in A&A

With a unique set of 54 overlapping narrow-band and two broader filters covering the entire optical range, the incoming Javalambre-Physics of the Accelerating Universe Astrophysical Survey (J-PAS) will provide a great opportunity for stellar physics and near-field cosmology. In this work, we use the miniJPAS data in 56 J-PAS filters and 4 complementary SDSS-like filters to explore and prove the potential of the J-PAS filter system in characterizing stars and deriving their atmospheric parameters. We obtain estimates for the effective temperature with a good precision (<150 K) from spectral energy distribution fitting. We have constructed the metallicity-dependent stellar loci in 59 colours for the miniJPAS FGK dwarf stars, after correcting certain systematic errors in flat-fielding. The very blue colours, including uJAVA-r, J0378-r, J0390-r, uJPAS-r, show the strongest metallicity dependence, around 0.25 mag/dex. The sensitivities decrease to about 0.1 mag/dex for the J0400-r, J0410-r, and J0420-r colours. The locus fitting residuals show peaks at the J0390, J0430, J0510, and J0520 filters, suggesting that individual elemental abundances such as [Ca/Fe], [C/Fe], and [Mg/Fe] can also be determined from the J-PAS photometry. Via stellar loci, we have achieved a typical metallicity precision of 0.1 dex. The miniJPAS filters also demonstrate strong potential in discriminating dwarfs and giants, particularly the J0520 and J0510 filters. Our results demonstrate the power of the J-PAS filter system in stellar parameter determinations and the huge potential of the coming J-PAS survey in stellar and Galactic studies., Comment: 16 pages, 16 figures, 3 tables, accepted by MNRAS

Recent results in image classification and extractive question answering have observed that pre-trained models trained on less in-distribution data have better out-of-distribution performance. However, it is unclear how broadly these trends hold. We conduct a large empirical study across three tasks, three broadly-applicable modeling interventions (increasing model size, using a different adaptation method, and pre-training on more data), and 14 diverse datasets to investigate the relationship between sample efficiency (amount of data needed to reach a given ID accuracy) and robustness (how models fare on OOD evaluation). We find that higher sample efficiency is only correlated with better average OOD robustness on some modeling interventions and tasks, but not others. On individual datasets, models with lower sample efficiency can even be more robust. These results suggest that general-purpose methods for improving sample efficiency are unlikely to yield universal OOD robustness improvements, since such improvements are highly dataset- and task-dependent. Even in an era of large, multi-purpose pretrained models, task-specific decisions may often be necessary for OOD generalization., Comment: 18 pages, 14 figures; to appear at ACL 2023

In this paper we analyze new observations from ALMA and VLA, at a high angular resolution corresponding to 5 - 8 au, of the protoplanetary disk around HD 163296 to determine the dust spatial distribution and grain properties. We fit the spectral energy distribution as a function of the radius at five wavelengths from 0.9 to 9\,mm, using a simple power law and a physical model based on an analytic description of radiative transfer that includes isothermal scattering. We considered eight dust populations and compared the models' performance using Bayesian evidence. Our analysis shows that the moderately high optical depth ($\tau$>1) at $\lambda \leq$ 1.3 mm in the dust rings artificially lower the millimeter spectral index, which should therefore not be considered as a reliable direct proxy of the dust properties and especially the grain size. We find that the outer disk is composed of small grains on the order of 200 $\mu$m with no significant difference between rings at 66 and 100 au and the adjacent gaps, while in the innermost 30 au, larger grains ($\geq$mm) could be present. We show that the assumptions on the dust composition have a strong impact on the derived surface densities and grain size. In particular, increasing the porosity of the grains to 80\% results in a total dust mass about five times higher with respect to grains with 25\% porosity. Finally, we find that the derived opacities as a function of frequency deviate from a simple power law and that grains with a lower porosity seem to better reproduce the observations of HD163296. While we do not find evidence of differential trapping in the rings of HD163296, our overall results are consistent with the postulated presence of giant planets affecting the dust temperature structure and surface density, and possibly originating a second-generation dust population of small grains., Comment: 27 pages, 28 figures, Accepted for publication in Astronomy and Astrophysics (A&A)

In the context of tensor network states, we for the first time reformulate the corner transfer matrix renormalization group (CTMRG) method into a variational bilevel optimization algorithm. The solution of the optimization problem corresponds to the fixed-point environment pursued in the conventional CTMRG method, from which the partition function of a classical statistical model, represented by an infinite tensor network, can be efficiently evaluated. The validity of this variational idea is demonstrated by the high-precision calculation of the residual entropy of the dimer model, and is further verified by investigating several typical phase transitions in classical spin models, where the obtained critical points and critical exponents all agree with the best known results in literature. Its extension to three-dimensional tensor networks or quantum lattice models is straightforward, as also discussed briefly., Comment: 7 pages, 7 figures

We discuss the prospects for improving the precision on the hadronic corrections to the anomalous magnetic moment of the muon, and the plans of the Muon $g-2$ Theory Initiative to update the Standard Model prediction., Comment: Contribution to the US Community Study on the Future of Particle Physics (Snowmass 2021)

Maximizing the discovery potential of increasingly precise neutrino experiments will require an improved theoretical understanding of neutrino-nucleus cross sections over a wide range of energies. Low-energy interactions are needed to reconstruct the energies of astrophysical neutrinos from supernovae bursts and search for new physics using increasingly precise measurement of coherent elastic neutrino scattering. Higher-energy interactions involve a variety of reaction mechanisms including quasi-elastic scattering, resonance production, and deep inelastic scattering that must all be included to reliably predict cross sections for energies relevant to DUNE and other accelerator neutrino experiments. This white paper discusses the theoretical status, challenges, required resources, and path forward for achieving precise predictions of neutrino-nucleus scattering and emphasizes the need for a coordinated theoretical effort involved lattice QCD, nuclear effective theories, phenomenological models of the transition region, and event generators., Comment: 81 pages, contribution to Snowmass 2021

Observations of black hole X-ray binaries and active galactic nuclei indicate that the accretion flows around black holes are composed of hot and cold gas, which have been theoretically described in terms of either a hot geometrically thick corona lying above and below a cold geometrically thin disk or an inner advection dominated accretion flow connected to an outer thin disk. This article reviews the accretion flows around black holes, with an emphasis on the physics that determines the configuration of hot and cold accreting gas, and how the configuration varies with the accretion rate and thereby produces various luminosity and spectra., Comment: 39 pages, 6 figures

The explosive growth of artificial intelligence and data-intensive computing has brought crucial challenge to modern information science and technology, i.e. conceptually new devices with superior properties are urgently desired. Memristor is recognized as a very promising circuit element to tackle the barriers, because of its fascinating advantages in imitating neural network of human brain, and thus realizing in-memory computing. However, there exist two core and fundamental issues: energy efficiency and accuracy, owing to the electric current operation of traditional memristors. In the present work, we demonstrate a new type of memristor, i.e. charge q and magnetic flux {\phi} space memristor, enabled by linear magnetoelectricity of Co4Nb2O9. The memory states show distinctly linear magnetoelectric coefficients with a large ratio of about 10, ensuing exceptional accuracy of related devices. The present q-{\phi} type memristor can be manipulated by magnetic and electric fields without involving electric current, paving the way to develop ultralow-energy-consuming devices. In the meanwhile, it is worth to mention that Co4Nb2O9 hosts an intrinsic compensated antiferromagnetic structure, which suggests interesting possibility of further integrating the unique merits of antiferromagnetic spintronics such as ultrahigh density and ultrafast switching. Linear magnetoelectricity is proposed to essential to the q-{\phi} type memristor, which would be accessible in a broad class of multiferroics and other magnetoelectric materials such as topological insulators. Our findings could therefore advance memristors towards new levels of functionality., Comment: 4 figures

For interpreting the behavior of a probabilistic model, it is useful to measure a model's calibration--the extent to which it produces reliable confidence scores. We address the open problem of calibration for tagging models with sparse tagsets, and recommend strategies to measure and reduce calibration error (CE) in such models. We show that several post-hoc recalibration techniques all reduce calibration error across the marginal distribution for two existing sequence taggers. Moreover, we propose tag frequency grouping (TFG) as a way to measure calibration error in different frequency bands. Further, recalibrating each group separately promotes a more equitable reduction of calibration error across the tag frequency spectrum.

Ferroelectricity, especially in hafnia-based thin films at nanosizes, has been rejuvenated in the fields of low-power, nonvolatile and Si-compatible modern memory and logic applications. Despite tremendous efforts to explore the formation of the metastable ferroelectric phase and the polarization degradation during field cycling, the ability of oxygen vacancy to exactly engineer and switch polarization remains to be elucidated. Here we report reversibly electrochemical control of ferroelectricity in Hf$_{0.5}$Zr$_{0.5}$O$_2$ (HZO) heterostructures with a mixed ionic-electronic LaSrMnO$_3$ electrode, achieving a hard breakdown field more than 18 MV/cm, over fourfold as high as that of typical HZO. The electrical extraction and insertion of oxygen into HZO is macroscopically characterized and atomically imaged in situ. Utilizing this reversible process, we achieved multiple polarization states and even repeatedly repaired the damaged ferroelectricity by reversed negative electric fields. Our study demonstrates the robust and switchable ferroelectricity in hafnia oxide distinctly associated with oxygen vacancy and opens up opportunities to recover, manipulate, and utilize rich ferroelectric functionalities for advanced ferroelectric functionality to empower the existing Si-based electronics such as multi-bit storage.

Dirac semimetals (DSMs) are classified into different phases based on the types of the Dirac fermions. Tuning the transition among different types of the Dirac fermions in one system remains challenging. Recently, KMgBi was predicted to be located at a critical state that various types of Dirac fermions can be induced owing to the existence of a flat band. Here, we carried out systematic studies on the electronic structure of KMgBi single crystal by combining angle-resolve photoemission spectroscopy (ARPES) and scanning tunneling microscopy/spectroscopy (STM/STS). The flat band was clearly observed near the Fermi level. We also revealed a small bandgap of ~ 20 meV between the flat band and the conduction band. These results demonstrate the critical state of KMgBi that transitions among various types of Dirac fermions can be tuned in one system., Comment: 15 pages, 4 figures

Topological Weyl semimetals (TWSs) are exotic crystals possessing emergent relativistic Weyl fermions connected by unique surface Fermi-arcs (SFAs) in their electronic structures. To realize the TWS state, certain symmetry (such as the inversion or time reversal symmetry) must be broken, leading to a topological phase transition (TPT). Despite the great importance in understanding the formation of TWSs and their unusual properties, direct observation of such a TPT has been challenging. Here, using a recently discovered magnetic TWS Co3Sn2S2, we were able to systematically study its TPT with detailed temperature dependence of the electronic structures by angle-resolved photoemission spectroscopy. The TPT with drastic band structures evolution was clearly observed across the Curie temperature (TC = 177 K), including the disappearance of the characteristic SFAs and the recombination of the spin-split bands that leads to the annihilation of Weyl points with opposite chirality. These results not only reveal important insights on the interplay between the magnetism and band topology in TWSs, but also provide a new method to control their exotic physical properties., Comment: 15 pages, 4 figures