arXiv subjects

Wei Zhu

Publications and source records attributed to Wei Zhu.

At least 145 records · Page 8Linked to original sources

arXiv2024

The Kinematic and Chemical Properties of the Close-in Planet Host Star 8 UMi

A recent study by Hon et al. reported that a close-in planet around the red clump star, 8 UMi, should have been engulfed during the expansion phase of its parent star's evolution. They explained the survival of this exoplanet through a binary-merger channel for 8 UMi. The key to testing this formation scenario is to derive the true age of this star: is it an old "imposter" resulting from a binary merger, or a genuinely young red clump giant? To accomplish this, we derive kinematic and chemical properties for 8 UMi using astrometric data from {\it Gaia} DR3 and the element-abundance pattern measured from a high-resolution ($R \sim 75,000$) spectrum taken by SOPHIE. Our analysis shows that 8 UMi is a normal thin-disk star with orbital rotation speed of $\it{V}_\mathrmϕ=\mathrm{244.96 km s^{-1}}$, and possesses a Solar metallicity ([Fe/H] $= -0.05 \pm 0.07$) and $α$-element abundance ratio ([$α$/Fe] $= +0.01 \pm 0.03$). By adopting well-established relationships between age and space velocities/elemental abundances, we estimate a kinematic age of $3.50^{+3.00}_{-2.00}$ Gyr, and a chemical age of $3.25^{+2.50}_{-1.50}$ Gyr from [C/N] and $3.47 \pm 1.96$ Gyr from [Y/Mg] for 8 UMi, respectively. These estimates are consistent with the isochrone-fitting age ($1.90^{+1.15}_{-0.30}$ Gyr) of 8 UMi, but are all much younger than the timescale required in a binary-merger scenario. This result challenges the binary-merger model; the existence of such a closely orbiting exoplanet around a giant star remains a mystery yet to be resolved.

astro-ph.EP↗

arXiv2024

ALoRA: Allocating Low-Rank Adaptation for Fine-tuning Large Language Models

Parameter-efficient fine-tuning (PEFT) is widely studied for its effectiveness and efficiency in the era of large language models. Low-rank adaptation (LoRA) has demonstrated commendable performance as a popular and representative method. However, it is implemented with a fixed intrinsic rank that might not be the ideal setting for the downstream tasks. Recognizing the need for more flexible downstream task adaptation, we extend the methodology of LoRA to an innovative approach we call allocating low-rank adaptation (ALoRA) that enables dynamic adjustments to the intrinsic rank during the adaptation process. First, we propose a novel method, AB-LoRA, that can effectively estimate the importance score of each LoRA rank. Second, guided by AB-LoRA, we gradually prune abundant and negatively impacting LoRA ranks and allocate the pruned LoRA budgets to important Transformer modules needing higher ranks. We have conducted experiments on various tasks, and the experimental results demonstrate that our ALoRA method can outperform the recent baselines with comparable tunable parameters.

cs.CL↗

arXiv2024

In-situ tunable giant electrical anisotropy in a grating gated AlGaN/GaN two-dimensional electron gas

Materials with in-plane electrical anisotropy have great potential for designing artificial synaptic devices. However, natural materials with strong intrinsic in-plane electrical anisotropy are rare. We introduce a simple strategy to produce extremely large electrical anisotropy via grating gating of a semiconductor two-dimensional electron gas (2DEG) of AlGaN/GaN. We show that periodically modulated electric potential in the 2DEG induces in-plane electrical anisotropy, which is significantly enhanced in a magnetic field, leading to an ultra large electrical anisotropy. This is induced by a giant positive magnetoresistance and a giant negative magnetoresistance under two orthogonally oriented in-plane current flows, respectively. This giant electrical anisotropy is in-situ tunable by tailoring both the grating gate voltage and the magnetic field. Our semiconductor device with controllable giant electrical anisotropy will stimulate new device applications, such as multi-terminal memtransistors and bionic synapses.

cond-mat.mes-hall↗

arXiv2024

The Metallicity Dimension of the Super Earth-Cold Jupiter Correlation

The correlation between close-in super Earths and distant cold Jupiters in planetary systems has important implications for their formation and evolution. In contrary to some earlier findings, a recent study conducted by Bonomo et al.\ suggests that the occurrence of cold Jupiter companions is not excessive in super Earth systems. Here we show that this discrepancy can be seen as a Simpson's paradox and is resolved once the metallicity dependence of the super Earth--cold Jupiter relation is taken into account. A common feature is noticed that almost all the cold Jupiter detections with inner super Earth companions are found around metal-rich stars. Focusing on the Sun-like hosts with super-solar metallicities, we show that the frequency of cold Jupiters conditioned on the presence of inner super Earths is $39_{-11}^{+12}\%$, whereas the frequency of cold Jupiters in the same metallicity range is no more than $20\%$. Therefore, the occurrences of close-in super Earths and distant cold Jupiters appear correlated around metal-rich hosts. The relation between the two types of planets remains unclear for stars with metal-poor hosts due to the limited sample size and the much lower occurrence rate of cold Jupiters, but a correlation between the two cannot be ruled out.

astro-ph.EP↗

arXiv2024

Continuum of spin excitations in an ordered magnet

Continuum of spin excitations observed in inelastic neutron scattering experiments are often considered as a strong evidence of quantum spin liquid formation. When quantum spin liquid is indeed the ground state of a disorder-free magnetic compound, the elementary excitation is no longer the conventional spin waves (magnons). Instead, the magnons fractionalize into spinons, leaving only a two-spinon continuum detectable in inelastic neutron scattering experiments. For a clean ordered antiferromagnet, it was unclear if we can observe a continuous spectrum similar to the ones in a quantum spin liquid state. Here we show that the magnetically ordered state in Na$_2$BaCo(PO$_4$)$_2$ is able to host a spin excitation continuum induced by strong quantum fluctuations. Thus, a second thought is necessary when concluding such continuum as signature of quantum spin liquid in new material explorations.

cond-mat.str-el↗

arXiv2024

Text2MDT: Extracting Medical Decision Trees from Medical Texts

Knowledge of the medical decision process, which can be modeled as medical decision trees (MDTs), is critical to build clinical decision support systems. However, the current MDT construction methods rely heavily on time-consuming and laborious manual annotation. In this work, we propose a novel task, Text2MDT, to explore the automatic extraction of MDTs from medical texts such as medical guidelines and textbooks. We normalize the form of the MDT and create an annotated Text-to-MDT dataset in Chinese with the participation of medical experts. We investigate two different methods for the Text2MDT tasks: (a) an end-to-end framework which only relies on a GPT style large language models (LLM) instruction tuning to generate all the node information and tree structures. (b) The pipeline framework which decomposes the Text2MDT task to three subtasks. Experiments on our Text2MDT dataset demonstrate that: (a) the end-to-end method basd on LLMs (7B parameters or larger) show promising results, and successfully outperform the pipeline methods. (b) The chain-of-thought (COT) prompting method \cite{Wei2022ChainOT} can improve the performance of the fine-tuned LLMs on the Text2MDT test set. (c) the lightweight pipelined method based on encoder-based pretrained models can perform comparably with LLMs with model complexity two magnititudes smaller. Our Text2MDT dataset is open-sourced at \url{https://tianchi.aliyun.com/dataset/95414}, and the source codes are open-sourced at \url{https://github.com/michael-wzhu/text2dt}.

cs.CL↗

arXiv2024

Efficient Multi-domain Text Recognition Deep Neural Network Parameterization with Residual Adapters

Recent advancements in deep neural networks have markedly enhanced the performance of computer vision tasks, yet the specialized nature of these networks often necessitates extensive data and high computational power. Addressing these requirements, this study presents a novel neural network model adept at optical character recognition (OCR) across diverse domains, leveraging the strengths of multi-task learning to improve efficiency and generalization. The model is designed to achieve rapid adaptation to new domains, maintain a compact size conducive to reduced computational resource demand, ensure high accuracy, retain knowledge from previous learning experiences, and allow for domain-specific performance improvements without the need to retrain entirely. Rigorous evaluation on open datasets has validated the model's ability to significantly lower the number of trainable parameters without sacrificing performance, indicating its potential as a scalable and adaptable solution in the field of computer vision, particularly for applications in optical text recognition.

cs.CV↗

arXiv2023

Overview of the PromptCBLUE Shared Task in CHIP2023

This paper presents an overview of the PromptCBLUE shared task (http://cips-chip.org.cn/2023/eval1) held in the CHIP-2023 Conference. This shared task reformualtes the CBLUE benchmark, and provide a good testbed for Chinese open-domain or medical-domain large language models (LLMs) in general medical natural language processing. Two different tracks are held: (a) prompt tuning track, investigating the multitask prompt tuning of LLMs, (b) probing the in-context learning capabilities of open-sourced LLMs. Many teams from both the industry and academia participated in the shared tasks, and the top teams achieved amazing test results. This paper describes the tasks, the datasets, evaluation metrics, and the top systems for both tasks. Finally, the paper summarizes the techniques and results of the evaluation of the various approaches explored by the participating teams.

cs.CL↗

arXiv2023

OGLE-2017-BLG-0448Lb: A Low Mass-Ratio Wide-Orbit Microlensing Planet?

The gravitational microlensing technique is most sensitive to planets in a Jupiter-like orbit and has detected more than 200 planets. However, only a few wide-orbit ($s > 2$) microlensing planets have been discovered, where $s$ is the planet-to-host separation normalized to the angular Einstein ring radius, $θ_{\rm E}$. Here we present the discovery and analysis of a strong candidate wide-orbit microlensing planet in the event, OGLE-2017-BLG-0448. The whole light curve exhibits long-term residuals to the static binary-lens single-source model, so we investigate the residuals by adding the microlensing parallax, microlensing xallarap, an additional lens, or an additional source. For the first time, we observe a complex degeneracy between all four effects. The wide-orbit models with $s \sim 2.5$ and a planet-to-host mass-ratio of $q \sim 10^{-4}$ are significantly preferred, but we cannot rule out the close models with $s \sim 0.35$ and $q \sim 10^{-3}$. A Bayesian analysis based on a Galactic model indicates that, despite the complicated degeneracy, the surviving wide-orbit models all contain a super-Earth-mass to Neptune-mass planet at a projected planet-host separation of $\sim 6$ au and the surviving close-orbit models all consist of a Jovian-mass planet at $\sim 1$ au. The host star is probably an M or K dwarf. We discuss the implications of this dimension-degeneracy disaster on microlensing light-curve analysis and its potential impact on statistical studies.

astro-ph.EP↗

arXiv2023

Gluon condensation: from nucleon to Galactic center

The Galactic Center Excess (GCE), one of the most remarkable discoveries by Fermi-LAT, has prompted extensive exploration over the past decade, often attributed to dark matter or millisecond pulsars. This work proposes a novel interpretation on the origin of the GCE, focusing on the observed spectral shape. Protons are accelerated at the Galactic center and collide with the neutron cluster on the surface of the non-rotating neutron stars. Due to the gluon condensation in nucleons, these collisions produce a large number of mesons, which have reached to the saturation state and subsequently generate the broken power law in the gamma ray spectra. We explained the spectral shape of GCE using the gluon condensation and an assumption of existing the non-rotating neutron stars at the Galactic center. This example of the gluon condensation mechanism not only expands the applications of the hadronic scenario in the cosmic gamma ray spectra but also provides a new evidence of the gluon condensation.

astro-ph.HE↗

arXiv2023

Inversion symmetry-broken tetralayer graphene probed by second harmonic generation

Symmetry breaking governs most fascinating phenomena in crystals, such as ferroelectricity, nonlinear optics, piezoelectricity, ferromagnetism, and superconductivity. In two-dimensional materials, a wide variety of tuning knobs presents extraordinary opportunities for engineering symmetry breaking, leading to the emergence and manipulation of novel physical properties. Recently, tetralayer graphene with ABCB stacking order is predicted to possess atypical elemental ferroelectricity arising from the symmetry breaking induced by its specific stacking configuration. Experimentally unveiling the stacking-order dependent symmetry in tetralayer graphene is crucial to understand the intricate properties in the emergent graphene allotropes. Here, we observe pronounced nonlinear optical second harmonic generation (SHG) in ABCB-stacked tetralayer graphene, but absent in both ABAB- and ABCA-stacked allotropes. Our results provide direct evidence of symmetry breaking in ABCB-stacked tetralayer graphene. The remarkable contrast in the SHG spectra of tetralayer graphene allows straightforward identification of ABCB domains from the other two kinds of stacking order and facilitates the characterization of their crystalline orientation. The employed SHG technique serves as a convenient tool for exploring the intriguing physics and novel nonlinear optics in ABCB-stacked graphene, where spontaneous polarization and intrinsic gapped flat bands coexist. Our results establish ABCB-stacked graphene as a unique platform for studying the rare ferroelectricity in non-centrosymmetric elemental structures.

cond-mat.mes-hall↗

arXiv2023

KMT-2023-BLG-1431Lb: A New $q < 10^{-4}$ Microlensing Planet from a Subtle Signature

The current studies of microlensing planets are limited by small number statistics. Follow-up observations of high-magnification microlensing events can efficiently form a statistical planetary sample. Since 2020, the Korea Microlensing Telescope Network (KMTNet) and the Las Cumbres Observatory (LCO) global network have been conducting a follow-up program for high-magnification KMTNet events. Here, we report the detection and analysis of a microlensing planetary event, KMT-2023-BLG-1431, for which the subtle (0.05 magnitude) and short-lived (5 hours) planetary signature was characterized by the follow-up from KMTNet and LCO. A binary-lens single-source (2L1S) analysis reveals a planet/host mass ratio of $q = (0.72 \pm 0.07) \times 10^{-4}$, and the single-lens binary-source (1L2S) model is excluded by $Δχ^2 = 80$. A Bayesian analysis using a Galactic model yields estimates of the host star mass of $M_{\rm host} = 0.57^{+0.33}_{-0.29}~M_\odot$, the planetary mass of $M_{\rm planet} = 13.5_{-6.8}^{+8.1}~M_{\oplus}$, and the lens distance of $D_{\rm L} = 6.9_{-1.7}^{+0.8}$ kpc. The projected planet-host separation of $a_\perp = 2.3_{-0.5}^{+0.5}$ au or $a_\perp = 3.2_{-0.8}^{+0.7}$, subject to the close/wide degeneracy. We also find that without the follow-up data, the survey-only data cannot break the degeneracy of central/resonant caustics and the degeneracy of 2L1S/1L2S models, showing the importance of follow-up observations for current microlensing surveys.

astro-ph.EP↗

arXiv2023

Systematic Reanalysis of KMTNet microlensing events, Paper I: Updates of the Photometry Pipeline and a New Planet Candidate

In this work, we update and develop algorithms for KMTNet tender-love care (TLC) photometry in order to create an new, mostly automated, TLC pipeline. We then start a project to systematically apply the new TLC pipeline to the historic KMTNet microlensing events, and search for buried planetary signals. We report the discovery of such a planet candidate in the microlensing event MOA-2019-BLG-421/KMT-2019-BLG-2991. The anomalous signal can be explained by either a planet around the lens star or the orbital motion of the source star. For the planetary interpretation, despite many degenerate solutions, the planet is most likely to be a Jovian planet orbiting an M or K dwarf, which is a typical microlensing planet. The discovery proves that the project can indeed increase the sensitivity of historic events and find previously undiscovered signals.

astro-ph.EP↗

arXiv2023

Planets Across Space and Time (PAST). V. The evolution of hot Jupiters revealed by the age distribution of their host stars

The unexpected discovery of hot Jupiters challenged the classical theory of planet formation inspired by our solar system. Until now, the origin and evolution of hot Jupiters are still uncertain. Determining their age distribution and temporal evolution can provide more clues into the mechanism of their formation and subsequent evolution. Using a sample of 383 giant planets around Sun-like stars collected from the kinematic catalogs of the Planets Across Space and Time (PAST) project, we find that hot Jupiters are preferentially hosted by relatively younger stars in the Galactic thin disk. We subsequently find that the frequency of hot Jupiters declines with age. In contrast, the frequency of warm/cold Jupiters shows no significant dependence on age. Such a trend is expected from the tidal evolution of hot Jupiters' orbits, and our result offers supporting evidence using a large sample. We also perform a joint analysis on the planet frequencies in the stellar age-metallicity plane. The result suggests that the frequencies of hot Jupiters and warm/cold Jupiters, after removing the age dependence are both correlated with stellar metallicities. Moreover, we show that the above correlations can explain the bulk of the discrepancy in hot Jupiter frequencies inferred from the transit and radial velocity (RV) surveys, given that RV targets tend to be more metal-rich and younger than transits.

astro-ph.EP↗

arXiv2023

Improving Prompt Tuning with Learned Prompting Layers

Prompt tuning prepends a soft prompt to the input embeddings or hidden states and only optimizes the prompt to adapt pretrained models (PTMs) to downstream tasks. The previous work manually selects prompt layers which are far from optimal and failed to exploit the potential of prompt tuning. In this work, we propose a novel framework, \underline{S}elective \underline{P}rompt \underline{T}uning (SPT), that learns to select the proper prompt layers by inserting a prompt controlled by a learnable probabilistic gate at each intermediate layer. We further propose a novel bi-level optimization framework, SPT-DARTS, that can better optimize the learnable gates and improve the final prompt tuning performances of the learned prompt layer settings. We conduct extensive experiments with ten benchmark datasets under the full-data and few-shot scenarios. The results demonstrate that our SPT framework can perform better than the previous state-of-the-art PETuning baselines with comparable or fewer tunable parameters.

cs.CL↗

arXiv2023

Uncovering conformal symmetry in the $3D$ Ising transition: State-operator correspondence from a fuzzy sphere regularization

The $3D$ Ising transition, the most celebrated and unsolved critical phenomenon in nature, has long been conjectured to have emergent conformal symmetry, similar to the case of the $2D$ Ising transition. Yet, the emergence of conformal invariance in the $3D$ Ising transition has rarely been explored directly, mainly due to unavoidable mathematical or conceptual obstructions. Here, we design an innovative way to study the quantum version of the $3D$ Ising phase transition on spherical geometry, using the "fuzzy (non-commutative) sphere" regularization. We accurately calculate and analyze the energy spectra at the transition, and explicitly demonstrate the state-operator correspondence (i.e. radial quantization), a fingerprint of conformal field theory. In particular, we have identified 13 parity-even primary operators within a high accuracy and 2 parity-odd operators that were not known before. Our result directly elucidates the emergent conformal symmetry of the $3D$ Ising transition, a conjecture made by Polyakov half a century ago. More importantly, our approach opens a new avenue for studying $3D$ CFTs by making use of the state-operator correspondence and spherical geometry.

cond-mat.stat-mech↗

arXiv2023

Sub-structure characteristic mode analysis of microstrip antennas using a global multi-trace formulation

A characteristic mode (CM) method that relies on a global multi-trace formulation (MTF) of surface integral equations is proposed to compute the modes and the resonance frequencies of microstrip patch antennas with finite dielectric substrates and ground planes. Compared to the coupled formulation of electric field and Poggio-Miller-Chang-Harrington-Wu-Tsai integral equations, global MTF allows for more direct implementation of a sub-structure CM method. This is achieved by representing the coupling of the electromagnetic fields on the substrate and ground plane in the form of a numerical Green function matrix, which yields a more compact generalized eigenvalue equation. The resulting sub-structure CM method avoids the cumbersome computation of the multilayered medium Green function (unlike the CM methods that rely on mixed-potential integral equations) and the volumetric discretization of the substrate (unlike the CM methods that rely on volume-surface integral equations), and numerical results show that it is a reliable and accurate approach to predicting the modal behavior of electromagnetic fields on practical microstrip antennas.

cs.CE↗

arXiv2023

PromptCBLUE: A Chinese Prompt Tuning Benchmark for the Medical Domain

Biomedical language understanding benchmarks are the driving forces for artificial intelligence applications with large language model (LLM) back-ends. However, most current benchmarks: (a) are limited to English which makes it challenging to replicate many of the successes in English for other languages, or (b) focus on knowledge probing of LLMs and neglect to evaluate how LLMs apply these knowledge to perform on a wide range of bio-medical tasks, or (c) have become a publicly available corpus and are leaked to LLMs during pre-training. To facilitate the research in medical LLMs, we re-build the Chinese Biomedical Language Understanding Evaluation (CBLUE) benchmark into a large scale prompt-tuning benchmark, PromptCBLUE. Our benchmark is a suitable test-bed and an online platform for evaluating Chinese LLMs' multi-task capabilities on a wide range bio-medical tasks including medical entity recognition, medical text classification, medical natural language inference, medical dialogue understanding and medical content/dialogue generation. To establish evaluation on these tasks, we have experimented and report the results with the current 9 Chinese LLMs fine-tuned with differtent fine-tuning techniques.

cs.CL↗