Search

Large Language Model-based Dense Retrieval (LLM-DR) optimizes over numerous heterogeneous fine-tuning collections from different domains. However, the discussion about its training data distribution is still minimal. Previous studies rely on empirically assigned dataset choices or sampling ratios, which inevitably leads to sub-optimal retrieval performances. In this paper, we propose a new task-level Distributionally Robust Optimization (tDRO) algorithm for LLM-DR fine-tuning, targeted at improving the universal domain generalization ability by end-to-end reweighting the data distribution of each task. The tDRO parameterizes the domain weights and updates them with scaled domain gradients. The optimized weights are then transferred to the LLM-DR fine-tuning to train more robust retrievers. Experiments show optimal improvements in large-scale retrieval benchmarks and reduce up to 30% dataset usage after applying our optimization algorithm with a series of different-sized LLM-DR models.

In this paper, we compute the total and differential cross sections for $e^+e^- \to J/\psi+c+\bar{c}$ at the $B$ factories up to next-to-leading order (NLO) corrections within the framework of nonrelativistic QCD factorization theory. We then obtain an improved pQCD prediction for this process by using the Principle of Maximum Conformality (PMC). We show the first time that the PMC can be applied for any pQCD calculable observable at the total or differential levels via a self-consistent way in perturbation theory. By the recursive use of renormalization group equation, the resultant PMC series satisfies all the requirements of renormalization group, which is free of conventional renormalization scheme-and-scale ambiguities at any fixed-order level. We thus obtain a more precise prompt total cross section at the NLO level, e.g. $\sigma|_{\rm prompt}^{\rm PMC}= 0.565^{+0.144}_{-0.125}~\text{pb}$. Here the uncertainty is the squared average of those from the $\alpha_s$ fixed-point uncertainty $\Delta\alpha_s(M_Z)$, the uncertainty of charm quark mass $\Delta m_c$, and an estimated contribution of the uncalculated NNLO-terms as predicted by the Pad\'{e} approximation approach. The differential cross sections $d\sigma/dP_{J/\psi}$, $d\sigma/d|\cos \theta|$, and $d\sigma/dz$ for $e^+e^- \to J/\psi+c+\bar{c}$ are further examined. Those results show that by further considering the feed-down contributions, the PMC predictions show better agreement with the Belle measurements., Comment: 22 pages, 11 figures

Scaling the size of a model enhances its capabilities but significantly increases computation complexity. Mixture-of-Experts models (MoE) address the issue by allowing model size to scale up without substantially increasing training or inference costs. In MoE, there is an important module called the router, which is used to distribute each token to the experts. Currently, the mainstream routing methods include dynamic routing and fixed routing. Despite their promising results, MoE models encounter several challenges. Primarily, for dynamic routing methods, the dispersion of training tokens across multiple experts can lead to underfitting, particularly for infrequent tokens. Additionally, though fixed routing methods can mitigate that issue, they compromise on the diversity of representations. In this paper, we propose \textbf{MaskMoE}, a method designed to enhance token-level learning by employing a routing \textbf{mask}ing technique within the \textbf{M}ixture-\textbf{o}f-\textbf{E}xperts model. MaskMoE is capable of maintaining representation diversity while achieving more comprehensive training. Experimental results demonstrate that our method outperforms previous dominant Mixture-of-Experts models in terms of both perplexity (PPL) and downstream task performance., Comment: Work in progress

We present a new approach for determining the strong coupling $\alpha_s(Q)$ over the entire perturbative range of validity, for scales from $\Lambda_{\mathrm{QCD}}$ up to the Planck scale ${\sim}10^{19}$\,GeV, with the highest precision and using the data of just a single experiment. The results obtained with this method are consistent with world averages and exhibit improved precision with respect to previous reports. \pacs{12.38.Bx, 13.66.Bc, 13.66.Jn, 13.87.-a,11.10.Gh, Comment: 7 pages, 4 figures

In this paper, we explore the properties of the bottom-quark on-shell mass ($M_b$) by using its relation to the $\overline{\rm MS}$ mass (${\overline m}_b$). At present, this $\overline{\rm MS}$-on-shell relation has been known up to four-loop QCD corrections, which however still has a $\sim 2\%$ scale uncertainty by taking the renormalization scale as ${\overline m}_b({\overline m}_b)$ and varying it within the usual range of $[{\overline m}_b({\overline m}_b)/2, 2 {\overline m}_b({\overline m}_b)]$. The principle of maximum conformality (PMC) has been adopted to achieve a more precise $\overline{\rm MS}$-on-shell relation by eliminating such scale uncertainty. As a step forward, we also estimate the magnitude of the uncalculated higher-order terms by using the Pad\'{e} approximation approach. Numerically, by using the $\overline{\rm MS}$ mass ${\overline m}_b({\overline m}_b)=4.18^{+0.03}_{-0.02}$ GeV as an input, our predicted value for the bottom-quark on-shell mass becomes $M_b\simeq 5.36^{+0.10}_{-0.07}$ GeV, where the uncertainty is the squared average of the ones caused by $\Delta \alpha_s(M_Z)$, $\Delta {\overline m}_b({\overline m}_b)$, and the estimated magnitude of the higher-order terms., Comment: 5 pages, 2 figures

In this paper, we study the quasi-two-body decays $B\to P f_0(500)\to P\pi^+\pi^-$ [with $P=(\pi, K, \eta, \eta^{\prime})$] within framework of perturbative QCD (PQCD) factorization approach. With the help of $\pi$-$\pi$ distribution amplitude and scalar form factor $F_{\pi\pi}(\omega^2)$, we calculate the $CP$ averaged branching fraction and the $CP$ asymmetry for the quasi-two-body decays $B\to P f_0(500)\to P\pi^+\pi^-$. Taking the quasi-two-body decay $B^+ \to \pi^+ f_0(500) \to \pi^+ \pi^+ \pi^-$ as an explicit example, we present the behavior of differential branching fraction and direct $CP$ violation versus the $\pi$-$\pi$ invariant mass. The total branching fraction and direct $CP$ violation are $\mathcal{B}(B^+\to \pi^+ [\sigma\to]\pi^+\pi^-) = (1.78 \pm 0.41\pm 0.51) \times 10^{-6}$ and $\mathcal{A}_{CP}(B^+\to \pi^+ [\sigma\to]\pi^+\pi^-) = (29.8\pm 11.1\pm 13.0)\%$ respectively. Our results could be tested by further experiments., Comment: 15 pages, 3 figures, matches published version

Large language models, initially pre-trained with a limited context length, can better handle longer texts by continuing training on a corpus with extended contexts. However, obtaining effective long-context data is challenging due to the scarcity and uneven distribution of long documents across different domains. To address this issue, we propose a Query-centric data synthesis method, abbreviated as Quest. Quest is an interpretable method based on the observation that documents retrieved by similar queries are relevant but low-redundant, thus well-suited for synthesizing long-context data. The method is also scalable and capable of constructing large amounts of long-context data. Using Quest, we synthesize a long-context dataset up to 128k context length, significantly outperforming other data synthesis methods on multiple long-context benchmark datasets. In addition, we further verify that the Quest method is predictable through scaling law experiments, making it a reliable solution for advancing long-context models.

In this paper, we carry out an investigation into the semileptonic decays $D_s^+ \to K^0\ell^+\nu_\ell$ with $\ell=(e,\mu)$ by employing the QCD light-cone sum rules approach. The vector transition form factor (TFF) $f_+^{D_s^+ K^0}(q^2)$ for $D_s^+\to K^0$ decay is calculated while considering its next-to-leading order contribution. Subsequently, we briefly introduce the twist-2, 3 kaon distribution amplitudes, which are calculated by using QCD sum rules within the framework of the background field theory. At the large recoil point, the TFF has $f_+^{D_s^+ K^0}(0)=0.692_{-0.026}^{+0.027}$. Then, we extrapolate $f_+^{D_s^+ K^0}(q^2)$ to the whole physical $q^2$-region via the simplified $z(q^2,t)$-series expansion, and the behavior of TFF $f_+^{D_s^+ K^0}(q^2)$ is exhibited in the numerical results part, including the theoretical and experimental predictions for comparison. In addition, we compute the differential branching fraction $\mathcal{B}(D_s^+ \to K^0\ell^+\nu_\ell)$ with the electron and muon channels, which are expected to be $\mathcal{B}(D_s^+ \to K^0e^+\nu_e)=3.379_{-0.275}^{+0.301}\times 10^{-3}$ and $\mathcal{B}(D_s^+ \to K^0\mu^+\nu_\mu)=3.351_{-0.273}^{+0.299}\times 10^{-3}$ as well as contained other results for comparison. Our results show good agreement with the BESIII measurements and theoretical predictions. Furthermore, we present our prediction with respect to the CKM matrix element $|V_{cd}|$ by using the $\mathcal{B}(D_s^+ \to K^0e^+\nu_e)$ result from BESIII Collaboration, yielding $|V_{cd}|=0.221_{-0.010}^{+0.008}$. Finally, we provide the ratio between $D_s^+ \to K^0e^+\nu_e$ and $D_s^+ \to \eta e^+\nu_e$ channels, i.e. $\mathcal{R}_{K^0/\eta}^e=0.144_{-0.020}^{+0.028}$., Comment: 5 pages, 3 figures, comments welcome

In this paper, we analyze the top-quark decay $t\to Wb$ up to next-to-next-to-next-to-leading order (N$^{3}$LO) QCD corrections. For the purpose, we first adopt the principle of maximum conformality (PMC) to deal with the initial pQCD series. Then we adopt the Bayesian analysis approach, which quantifies the unknown higher-order terms' contributions in terms of a probability distribution, to estimate the possible magnitude of the uncalculated N$^{4}$LO-terms. In our calculation, an effective strong coupling constant $\alpha_{s}(Q_{*})$ is determined by using all non-conformal $\{\beta_{i}\}$ terms associated with the renormalization group equation. This leads to a next-to-leading-log PMC scale $Q_{*}^{(\rm NLL)}=10.3048$ GeV, which can be regarded as the correct momentum flow of the process. Consequently, we obtain an improved scale-invariant pQCD prediction for the top-quark decay width, e.g. $\Gamma_{t}^{\rm tot} = 1.3120 \pm 0.0038$ GeV, whose error is the squared average of the uncertainties from the decay width of $W$-boson $\Delta \Gamma_{W} = \pm 0.042$ GeV, the coupling constant $\Delta \alpha_{s}(m_{Z}) = \pm 0.0009$, and the predicted N$^{4}$LO-terms. The magnitude of the top-quark pole mass greatly affects the total decay width. By further taking the PDG top-quark pole mass error from cross-section measurements into consideration, e.g. $\Delta m_{t} = \pm 0.7$ GeV, we obtain $\Gamma_{t}^{\rm tot} = 1.3120 ^{+0.0194}_{-0.0192}$ GeV., Comment: 9 pages, 4 figures

In this paper, we present an improved calculation for the decay rate of the rare $Z$-boson decay into $J/\psi + J/\psi$. This decay is dominated by the photon fragmentation mechanism, i.e., the transition $Z\to J/\psi + \gamma^{*}$ followed by the fragmentation $\gamma^{*}\to J/\psi$. In our calculation, the amplitude of $\gamma^{*}\to J/\psi$ is extracted from the measured value of $\Gamma(J/\psi \to e^+ e^-)$, and the amplitude of $Z\to J/\psi + \gamma^{*}$ is calculate through the light-cone approach. The higher-order QCD and relativistic corrections in the amplitude of $\gamma^{*}\to J/\psi$ and the large logarithms of $m_{_Z}^2/m_c^2$ that appear in the amplitude of $Z\to J/\psi + \gamma^{*}$ are resummed in our calculation. Besides, the non-fragmentation amplitude is calculated based on the NRQCD factorization, and the next-to-leading order QCD and relativistic corrections are included. The obtained branching fraction for this $Z$ decay channel is $8.66 ^{+1.48} _{-0.69}\times 10^{-11}$., Comment: 10 pages, 2 figures

In the present paper, we study the properties of $\phi$-meson longitudinal leading-twist light-cone distribution amplitude $\phi_{2;{\phi}}^{\|}(x,\mu)$ by starting from a light-cone harmonic oscillator model for its wavefunction. To fix the input parameters, we derive the first ten $\xi$-moments of $\phi_{2;{\phi}}^{\|}(x,\mu)$ by using the QCD sum rules approach under the background field theory. The shape of $\phi_{2;{\phi}}^{\|}(x,\mu=2~{\rm GeV})$ tends to be a single-peak behavior, which is consistent with the latest Lattice QCD result. As an application, we derive the $D^+_s \to \phi$ transition form factors (TFFs) by using the light-cone sum rules approach. At the large recoil point, we obtain $A_1(0) = 0.512_{-0.020}^{+0.030}$, $A_2(0) = 0.402_{-0.067}^{+0.078}$, $A_0(0) = 0.596_{-0.020}^{+0.025}$ and $V(0) = 0.882_{-0.036}^{+0.040}$. As for the two typical ratios $\gamma_V$ and $\gamma_2$, we obtain $\gamma_V = 1.723_{-0.021}^{+0.023}$ and $\gamma_2 = 0.785_{-0.104}^{+0.100}$. After extrapolating those TFFs to the physically allowable region, we then obtain the transverse, longitudinal and total decay widths for semi-leptonic decay $D^+_s\to\phi\ell^+\nu_{\ell}$. Then the branching fractions are ${\cal B}(D^+_s\to \phi e^+\nu_e) = (2.367_{-0.132}^{+0.256})\times 10^{-3}$ and ${\cal B}(D^+_s\to \phi \mu^+\nu_{\mu}) = (2.349_{-0.132}^{+0.255})\times 10^{-3}$, which show good agreement with the data issued by the BESIII, the CLEO, and the BABAR Collaborations. We finally calculate $D^+_s\to\phi\ell^+ \nu_\ell$ polarization and asymmetry parameters., Comment: 32 pages, 7 figures, comments welcome

For the Fornberg-Whitham equation, the local well-posedness in the critical Besov space $B_{p, 1}^{1+\frac{1}{p}}(\mathbb{R})$ with $1\leq p <\infty$ has been studied in (Guo, Nonlinear Anal. RWA., 2023). However, for the endpoint case $p=\infty$, whether it is locally well-posed or ill-posed in $B_{\infty, 1}^{1}(\mathbb{R})$ is still unknown. In this paper, we prove that the Fornberg-Whitham equation is well-posed in the critical Besov space $B_{\infty, 1}^{1}(\mathbb{R})$ with solutions depending continuously on initial data, which is different from that of the Camassa-Holm equation (Guo et al., J. Differ. Equ., 2022). In addition, we show that this dependence is sharp by showing that the solution map is not uniformly continuous on the initial data.

In this paper, we reanalyze the top-quark pair production at the next-to-next-to-leading order (NNLO) in QCD at future $e^+e^-$ colliders by using the Principle of Maximum Conformality (PMC) method. The PMC renormalization scales in $\alpha_s$ are determined by absorbing the non-conformal $\beta$ terms by recursively using the Renormalization Group Equation (RGE). Unlike the conventional scale-setting method of fixing the scale at the center-of-mass energy $\mu_r=\sqrt{s}$, the determined PMC scale $Q_\star$ is far smaller than the $\sqrt{s}$ and increases with the $\sqrt{s}$, yielding the correct physical behavior for the top-quark pair production process. Moreover, the convergence of the pQCD series for the top-quark pair production is greatly improved due to the elimination of the renormalon divergence. For a typical collision energy of $\sqrt{s}=500$ GeV, the PMC scale is $Q_\star=107$ GeV; the QCD correction factor $K$ for conventional results is $K\sim1+0.1244^{+0.0102+0.0012}_{-0.0087-0.0011}+0.0184^{-0.0086+0.0002}_{+0.0061-0.0003}$, where the first error is caused by varying the scale $\mu_r\in[\sqrt{s}/2, 2\sqrt{s}]$ and the second error is from the top-quark mass $\Delta{m_t}=\pm0.7$ GeV. After applying the PMC, the renormalization scale uncertainty is eliminated and the QCD correction factor $K$ is improved to $K\sim 1+0.1507^{+0.0015}_{-0.0015}-0.0057^{+0.0001}_{-0.0000}$, where the error is from the top-quark mass $\Delta{m_t}=\pm0.7$ GeV. The PMC improved predictions for the top-quark pair production are helpful for detailed studies of properties of the top-quark at future $e^+e^-$ colliders., Comment: 8 pages, 3 figures. Accepted by Chinese Physics C

Masked auto-encoder pre-training has emerged as a prevalent technique for initializing and enhancing dense retrieval systems. It generally utilizes additional Transformer decoder blocks to provide sustainable supervision signals and compress contextual information into dense representations. However, the underlying reasons for the effectiveness of such a pre-training technique remain unclear. The usage of additional Transformer-based decoders also incurs significant computational costs. In this study, we aim to shed light on this issue by revealing that masked auto-encoder (MAE) pre-training with enhanced decoding significantly improves the term coverage of input tokens in dense representations, compared to vanilla BERT checkpoints. Building upon this observation, we propose a modification to the traditional MAE by replacing the decoder of a masked auto-encoder with a completely simplified Bag-of-Word prediction task. This modification enables the efficient compression of lexical signals into dense representations through unsupervised pre-training. Remarkably, our proposed method achieves state-of-the-art retrieval performance on several large-scale retrieval benchmarks without requiring any additional parameters, which provides a 67% training speed-up compared to standard masked auto-encoder pre-training with enhanced decoding., Comment: Accepted by SIGIR24. Our code is available at https://github.com/ma787639046/bowdpr

In this paper, we present an improved analysis of the top-quark pair production via the process $e^{+}e^{-}\to \gamma^{*}\to t\bar{t}$ near the threshold region up to next-to-next-to-next-to-leading order (N$^3$LO) QCD corrections. Near the threshold region, the top-quark velocity $v$ tends to zero, leading to Coulomb singularity. To achieve a reasonable prediction in the threshold region, we reconstruct the analytical expression for the Coulomb-terms up to N$^{3}$LO accuracy by using the PSLQ algorithm, whose numerical values agree well with the previous N$^3$LO-level calculations. It is found that the N$^{3}$LO series still has sizable renormalization scale dependence, and to improve the precision of the series, we apply the Principle of Maximum Conformality to eliminate such scale dependence. After that, the Coulomb part is resummed into a Sommerfeld-Gamow-Sakharov factor, which finally leads to a much more reasonable behavior near the threshold region., Comment: 10 pages, 6 figures

In this paper, we calculate the decay widths and branching fractions for the decays $Z \to H+ \gamma$ up to ${\cal O}(\alpha_{s} v^2)$ accuracy within the framework of nonrelativistic QCD, where $H$ stands for the $S$-wave quarkonium $\eta_c$, $J/\psi$, $\eta_b$ or $\Upsilon$, respectively. To compare with the leading-order terms, those corrections show good perturbative behavior as expected. It is found that contributions from the next-to-leading order QCD correction ${\cal O}(\alpha_{s}v^0)$, the relativistic correction ${\cal O}(\alpha^{0}_{s}v^2)$ and their joint correction ${\cal O}(\alpha_{s} v^2)$ are sizable and comparable to each other, especially for the charmonium case. Thus we need to take all of them into consideration for a sound estimation. For a high luminosity electron-positron collider running around the $Z$-pole, due to $Z$-boson resonance effect, sizable events could be produced from those rare decay channels., Comment: 8 pages, 2 figures

The precision and predictive power of perturbative QCD (pQCD) prediction depends on both a precise, convergent fixed-order series and a reliable way of estimating the contributions of unknown higher-order (UHO) terms. It has been shown that by applying the Principal of Maximum Conformality (PMC), which applies the renormalization group equation recursively to set the effective magnitude of $\alpha_s$ of the process, the remaining conformal coefficients will be well matched with the corresponding $\alpha_s$ at each orders, leading to a scheme-and-scale invariant and convergent perturbative series. Thus different from conventional scheme-and-scale dependent fixed-order series, the PMC series will provide a more reliable platform for estimating UHO contributions. In this paper, by using the total decay width $\Gamma(H\to\gamma\gamma)$ which has been calculated up to N$^4$LO QCD corrections, we derive its PMC series by using the PMC single-scale setting approach and estimate its unknown N$^5$LO contributions by using the Bayesian analysis. The Bayesian-based approach estimates the magnitude of the UHO contributions based on an optimized analysis of probability density distribution, and the predicted UHO contribution becomes more accurate when more loop terms have been known to tame the probability density function. Using the top-quark pole mass $M_t$=172.69 GeV and the Higgs mass $M_H$=125.25 GeV as inputs, we obtain $\Gamma(H\to\gamma\gamma) =9.56504~{\rm keV}$ and the estimated N$^5$LO contribution to the total decay width is $\Delta\Gamma_H=\pm1.65\times10^{-4}~{\rm keV}$ for the smallest credible interval of $95.5\%$ degree-of-belief., Comment: 9 pages, 2 figures

Generative language models are usually pretrained on large text corpus via predicting the next token (i.e., sub-word/word/phrase) given the previous ones. Recent works have demonstrated the impressive performance of large generative language models on downstream tasks. However, existing generative language models generally neglect an inherent challenge in text corpus during training, i.e., the imbalance between frequent tokens and infrequent ones. It can lead a language model to be dominated by common and easy-to-learn tokens, thereby overlooking the infrequent and difficult-to-learn ones. To alleviate that, we propose a MiLe Loss function for mitigating the bias of learning difficulties with tokens. During training, it can dynamically assess the learning difficulty of a to-be-learned token, according to the information entropy of the corresponding predicted probability distribution over the vocabulary. Then it scales the training loss adaptively, trying to lead the model to focus more on the difficult-to-learn tokens. On the Pile dataset, we train generative language models at different scales of 468M, 1.2B, and 6.7B parameters. Experiments reveal that models incorporating the proposed MiLe Loss can gain consistent performance improvement on downstream benchmarks., Comment: This paper has been accepted by NAACL 2024

The photoprodution of doubly heavy baryon ($\Xi_{cc},\Xi_{bb},\Xi_{bc}$) is investigated in the context of future high-energy and high-luminosity $e^+e^-$ colliders. The study incorporates two sources of initial photons, namely the LBS photon and the WWA photon. Alongside the direct photoproduction via the sub-process $\gamma+\gamma \rightarrow \Xi_{QQ^{'}} +\bar{Q}+\bar{Q^{'}}$ ($Q^{(')}=c,b$), the resolved photoproduction channels are specifically considered, encompassing the sub-processes $\gamma + g \rightarrow \Xi_{QQ^{'}} +\bar{Q}+\bar{Q^{'}}$, $g + g \rightarrow \Xi_{QQ^{'}} +\bar{Q}+\bar{Q^{'}}$, and $q + \bar{q} \rightarrow \Xi_{QQ^{'}} +\bar{Q}+\bar{Q^{'}}$ with $q=u,d,s$. Within the framework of non-relativistic QCD, two $(cc(bb))$-diquark configurations, ${}_{\bar{\textbf{3}}}[{}^3S_1]$ and ${}_{\textbf{6}}[{}^1S_0]$, and four $(bc)$-diquark configurations, $(bc)_{\bar{\textbf{3}}}[{}^3S_1]$, $(bc)_{\textbf{6}}[{}^1S_0]$, $(bc)_{\textbf{6}}[{}^3S_1]$ and $(bc)_{\bar{\textbf{3}}}[{}^1S_0]$, are considered in the calculations. Numerical results show that the single resolved photoproduction processes provide dominant contributions under certain collision configuration. At the future $e^+e^-$ colliders, the doubly heavy baryon generated via the photoproduction mechanism is promisingly observable and can be well studied.

In this paper, we focus on zero-filter limit problem for the Camassa-Holm equation in the more general Besov spaces. We prove that the solution of the Camassa-Holm equation converges strongly in $L^\infty(0,T;B^s_{2,r}(\R))$ to the inviscid Burgers equation as the filter parameter $\alpha$ tends to zero with the given initial data $u_0\in B^s_{2,r}(\R)$. Moreover, we also show that the zero-filter limit for the Camassa-Holm equation does not converges uniformly with respect to the initial data in $B^s_{2,r}(\R)$.

We present results for the thrust distribution in the electron positron annihilation to the three jet process at NNLO in the perturbative conformal window of QCD, as a function of the number of flavors $N_f$. Given the existence of an infrared interacting fixed point in this region, we can compare the Conventional Scale Setting (CSS) and the Principle of Maximum Conformality (PMC$_\infty$) methods along the entire renormalization group flow from the highest energies to zero energy. We then consider also the QED thrust, obtained as the limit $N_c \rightarrow 0$ of the number of colors and we show analogous comparison. QED in the low energy regime develops an infrared non-interacting fixed point. Using these quantum field theory limits as theoretical laboratories, we arrive at interesting results showing new features of the PMC$_\infty$., Comment: "Submission to SciPost", 14pages, 5 figures, proceedings , conference RADCOR 2023