arXiv subjects

Hao Zou

Publications and source records attributed to Hao Zou.

53 records · Page 3Linked to original sources

arXiv2022

Improving Question Answering with Generation of NQ-like Questions

Question Answering (QA) systems require a large amount of annotated data which is costly and time-consuming to gather. Converting datasets of existing QA benchmarks are challenging due to different formats and complexities. To address these issues, we propose an algorithm to automatically generate shorter questions resembling day-to-day human communication in the Natural Questions (NQ) dataset from longer trivia questions in Quizbowl (QB) dataset by leveraging conversion in style among the datasets. This provides an automated way to generate more data for our QA systems. To ensure quality as well as quantity of data, we detect and remove ill-formed questions using a neural classifier. We demonstrate that in a low resource setting, using the generated data improves the QA performance over the baseline system on both NQ and QB data. Our algorithm improves the scalability of training data while maintaining quality of data for QA systems.

cs.CL↗

arXiv2022

Quantum cohomology of symplectic flag manifolds

We compute the quantum cohomology of symplectic flag manifolds. Symplectic flag manifolds can be described by non-abelian GLSMs with superpotential. Although the ring relations cannot be directly read off from the equations of motion on the Coulomb branch due to complication introduced by the non-abelian gauge symmetry, it can be shown that they can be extracted from the localization formula in a gauge-invariant form. Our result is general for all symplectic flag manifolds, which reduces to previously established results on symplectic Grassmannians and complete symplectic flag manifolds derived by other means. We also explain why a (0,2) deformation of the GLSM does not give rise to a deformation of the quantum cohomology.

hep-th↗

arXiv2022

Remarks on 2d Unframed Quiver Gauge Theories

We study 2d $\mathcal{N}=(2,2)$ quiver gauge theories without flavor nodes. There is a special class of quivers whose gauge group ranks stay positive in any duality frame. We illustrate this with the Abelian Kronecker quiver and the Abelian Markov quiver as the simplest examples. In the geometric phase, they engineer an infinite sequence of projective spaces and hypersurfaces in Calabi-Yau spaces, respectively. We show that the Markov quiver provides an Abelianization of SU(3) SQCD. Turning on the FI parameters and the $θ$ angles for the Abelian quiver effectively deform SQCD by such parameters. For an Abelian necklace quiver corresponding to SU($k$) SQCD, we find evidence for singular loci supporting non-compact Coulomb branches in the Kähler moduli space.

hep-th↗

arXiv2022

Applying machine learning to the Calabi-Yau orientifolds with string vacua

We use the machine learning technique to search the polytope which can result in an orientifold Calabi-Yau hypersurface and the "naive Type IIB string vacua". We show that neural networks can be trained to give a high accuracy for classifying the orientifold property and vacua based on the newly generated orientifold Calabi-Yau database with $h^{1,1}(X) \leq 6$ arXiv:2111.03078. This indicates the orientifold symmetry may already be encoded in the polytope structure. In the end, we try to use the trained neural networks model to go beyond the database and predict the orientifold signal of polytope for higher $h^{1,1}(X)$.

hep-th↗

arXiv2022

CausPref: Causal Preference Learning for Out-of-Distribution Recommendation

In spite of the tremendous development of recommender system owing to the progressive capability of machine learning recently, the current recommender system is still vulnerable to the distribution shift of users and items in realistic scenarios, leading to the sharp decline of performance in testing environments. It is even more severe in many common applications where only the implicit feedback from sparse data is available. Hence, it is crucial to promote the performance stability of recommendation method in different environments. In this work, we first make a thorough analysis of implicit recommendation problem from the viewpoint of out-of-distribution (OOD) generalization. Then under the guidance of our theoretical analysis, we propose to incorporate the recommendation-specific DAG learner into a novel causal preference-based recommendation framework named CausPref, mainly consisting of causal learning of invariant user preference and anti-preference negative sampling to deal with implicit feedback. Extensive experimental results from real-world datasets clearly demonstrate that our approach surpasses the benchmark models significantly under types of out-of-distribution settings, and show its impressive interpretability.

cs.LG↗

arXiv2021

Semantic Segmentation-assisted Scene Completion for LiDAR Point Clouds

Outdoor scene completion is a challenging issue in 3D scene understanding, which plays an important role in intelligent robotics and autonomous driving. Due to the sparsity of LiDAR acquisition, it is far more complex for 3D scene completion and semantic segmentation. Since semantic features can provide constraints and semantic priors for completion tasks, the relationship between them is worth exploring. Therefore, we propose an end-to-end semantic segmentation-assisted scene completion network, including a 2D completion branch and a 3D semantic segmentation branch. Specifically, the network takes a raw point cloud as input, and merges the features from the segmentation branch into the completion branch hierarchically to provide semantic information. By adopting BEV representation and 3D sparse convolution, we can benefit from the lower operand while maintaining effective expression. Besides, the decoder of the segmentation branch is used as an auxiliary, which can be discarded in the inference stage to save computational consumption. Extensive experiments demonstrate that our method achieves competitive performance on SemanticKITTI dataset with low latency. Code and models will be released at https://github.com/jokester-zzz/SSA-SC.

cs.CV↗

arXiv2020

F-Siamese Tracker: A Frustum-based Double Siamese Network for 3D Single Object Tracking

This paper presents F-Siamese Tracker, a novel approach for single object tracking prominently characterized by more robustly integrating 2D and 3D information to reduce redundant search space. A main challenge in 3D single object tracking is how to reduce search space for generating appropriate 3D candidates. Instead of solely relying on 3D proposals, firstly, our method leverages the Siamese network applied on RGB images to produce 2D region proposals which are then extruded into 3D viewing frustums. Besides, we perform an online accuracy validation on the 3D frustum to generate refined point cloud searching space, which can be embedded directly into the existing 3D tracking backbone. For efficiency, our approach gains better performance with fewer candidates by reducing search space. In addition, benefited from introducing the online accuracy validation, for occasional cases with strong occlusions or very sparse points, our approach can still achieve high precision, even when the 2D Siamese tracker loses the target. This approach allows us to set a new state-of-the-art in 3D single object tracking by a significant margin on a sparse outdoor dataset (KITTI tracking). Moreover, experiments on 2D single object tracking show that our framework boosts 2D tracking performance as well.

cs.CV↗

arXiv2019

Supersymmetric Localization in GLSMs for Supermanifolds

In this paper we apply supersymmetric localization to study gauged linear sigma models (GLSMs) describing supermanifold target spaces. We use the localization method to show that A-twisted GLSM correlation functions for certain supermanifolds are equivalent to A-twisted GLSM correlation functions for hypersurfaces in ordinary spaces under certain conditions. We also argue that physical two-sphere partition functions are the same for these two types of target spaces. Therefore, we reproduce the claim of arXiv:hep-th/9404186, arXiv:hep-th/9506070. Furthermore, we explore elliptic genera and (0,2) deformations and find similar phenomena.

hep-th↗

arXiv2015

Effective Superpotentials of Type II D-brane/F-theory on Compact Complete Intersection Calabi-Yau Threefolds

In this paper, we extend the GKZ-system method to the more general case: compact Complete Intersection Calabi-Yau manifolds (CICY). For several one-deformation modulus compact CICYs with D-branes, the on-shell superpotentials in this paper from the extended GKZ-system method are exactly consistent with published results obtained from other methods. We further compute the off-shell superpotentials of these models. Then we obtain both the on-shell and off-shell superpotentials for sev- eral two-deformation moduli compact CICYs with D-branes by using the extended GKZ-system method. The discrete symmetrical groups, Z2, Z3 and Z4, of the holo- morphic curves wrapped by D-branes play the important roles in computing the superpotentials, in some sense, they are the quantum symmetries of these models. Furthermore, through the mirror symmetry, the Ooguri-Vafa invariants are extracted from the A-model instanton expansion.

hep-th↗

arXiv2011

Next-to-leading-order correction to pion form factor in $k_T$ factorization

We calculate next-to-leading-order (NLO) correction to the pion electromagnetic form factor at leading twist in the $k_T$ factorization theorem. Partons off-shell by $k_T^2$ are considered in both quark diagrams and effective diagrams for the transverse-momentum-dependent (TMD) pion wave function. The light-cone singularities in the TMD pion wave function are regularized by rotating the Wilson lines away from the light cone. The soft divergences from gluon exchanges among initial- and final-state partons cancel exactly. We derive the infrared-finite $k_T$-dependent NLO hard kernel for the pion electromagnetic form factor by taking the difference of the above two sets of diagrams. Varying the renormalization and factorization scales, we find that the NLO correction is smaller, when both the scales are set to the invariant masses of internal particles: it becomes lower than 40% of the leading-order (LO) contribution for momentum transfer squared $Q^2>7$ GeV$^2$. It is observed that the NLO leading-twist correction does not play an essential role in explaining the experimental data, but the LO higher-twist contribution does.

hep-ph↗

arXiv2009

The CKM suppressed $B(B_s) \to \bar D_{(s)}P,\bar D_{(s)}V,\bar D_{(s)}^P, \bar D_{(s)}^V$ decays in perturbative QCD approach

Although the two-body charmed decays $B_{(s)} \to \bar D_{(s)}^{(*)}P$ and $\bar D_{(s)}^{(*)}V$, where $P$($V$) denotes a light pseudoscalar(vector) meson, are CKM suppressed comparing with the $B_{(s)} \to D_{(s)}^{(*)}P$ and $ D_{(s)}^{(*)}V$ decays, they are important in the CKM angle $γ=ϕ_3$ extraction method. We investigated these decays in the perturbative QCD approach to the leading order of $m_D/m_B$ and $Λ_{\rm{QCD}}/m_D$ expansion. We find that the nonfactorizable emission diagrams and the annihilation diagrams are not negligible in many of these channels. The numerical results show that most channels have branching ratios with an order of $10^{-6}$ or $10^{-7}$. The ratio needed for the CKM angle $γ$ extraction is estimated as $r = \frac{|A(B^- \to \bar{D}^0 K^-)|}{|A(B^- \to D^0 K^-)|} = 0.092_{-0.003-0.003}^{+0.012+0.003},$ which is too small for the experiments. Some of the $ B_{(s)}\to \bar D_{(s)}^*V$ decays have a very large transversely polarized contribution that can reach 80%.

hep-ph↗

arXiv2009

Anatomy of the pQCD Approach to the Baryonic Decays $Λ_b \to pπ, p K$

We calculate the CP-averaged branching ratios and CP-violating asymmetries for the two-body charmless hadronic decays $Λ_b \to p π, pK$ in the perturbative QCD (pQCD) approach to lowest order in $α_s$. The baryon distribution amplitudes involved in the factorization formulae are considered to the leading twist accuracy and the distribution amplitudes of the proton are expanded to the next-to-leading conformal spin (i.e., "P" -waves), the moments of which are determined from QCD sum rules. Our work shows that the contributions from the factorizable diagrams in $Λ_b \to p π, pK$ decays are much smaller compared to the non-factorizable diagrams in the conventional pQCD approach. We argue that this reflects the estimates of the $Λ_b \to p$ transition form factors in the $k_T$ factorization approach, which are found typically an order of magnitude smaller than those estimated in the light-cone sum rules and in the non-relativistic quark model. As an alternative, we adopt a hybrid pQCD approach, in which we compute the factorizable contributions with the $Λ_b \to p$ form factors taken from the light cone QCD sum rules. The non-factorizable diagrams are evaluated utilizing the conventional pQCD formalism which is free from the endpoint singularities. The predictions worked out here are confronted with the recently available data from the CDF collaboration on the branching ratios and the direct CP asymmetries for the decays $ Λ_b \to p π$, and $Λ_b \to p K$. The asymmetry parameter $α$ relevant for the anisotropic angular distribution of the emitted proton in the polarized $Λ_b$ baryon decays is also calculated for the two decay modes.

hep-ph↗

arXiv2008

FCNC-induced semileptonic decays of $J/ψ$ in the Standard Model

In this work, we calculate the form factors for $J/ψ\to \bar{D}^{(*)0}$ induced by the flavor changing neutral currents (FCNC) in terms of the QCD sum rules. Making use of these form factors, we further calculate the branching fractions of semileptonic decays $J/ψ\to \bar{D}^{(*)0} l ^+ l^-$ ($l=e, μ$). In particular, we formulate the matrix element $ $ with $T_{μν}$ being a tensor current, which was not fully discussed in previous literature. Our analysis indicates that if only the standard model (SM) applies, the production of single charmed mesons at the present electron-positron colliders is too small to be observed even the resonance effects are included, therefore if an anomalous production rates are observed, it would be a hint of new physics beyond SM. Even though the predicted branching ratios are beyond the reach of present facilities which can be seen from a rough order estimate, the more accurate formulation of the three point correlation function derived in this work has theoretical significance and the technique can also be applied to other places. In analog to some complicated theoretical derivations which do not have immediate phenomenological application yet, if the future experiments can provide sufficient luminosity and accuracy, the results would be helpful.

hep-ph↗

arXiv2008

The $B(B_s)\to D_{(s)} P$, $D_{(s)} V$, $D_{(s)}^{}P$ and $D_{(s)}^{}V$ decays in the perturbative QCD approach

Two-body non-leptonic charmed decays $B_{(s)} \to D_{(s)}P$, $D_{(s)}^*P$, $D_{(s)}V$ and $D_{(s)}^*V$ are analyzed in perturbative QCD approach, where $P$ and $V$ denote the light pseudoscalar meson and vector meson, respectively. We test the $D$ meson wave function by a $χ^2$ fit with experimental data of six $B\to DP$ channels. We give the branching ratios of all the charmed B decay channels, most of which agree with experiments amazingly well. The predicted $B_s$ decays can be confronted with the future experimental data. By straightforward calculations, our pQCD approach gives the right relative strong phase of $a_2/a_1$ with experiments. We also predict the percentage of transverse polarizations in $B_{(s)} \to D^* V$ decay channels.

hep-ph↗

arXiv2008

Weak decays of $J/ψ$: the non-leptonic case

In our previous study, we calculated the transition from factors of $J/ψ\to D^{(*)}_{(s)}$ using the QCD sum rules. Based on the factorization approximation, the obtained form factors can be applied to evaluate the weak non-leptonic decay rates of $J/ψ\to D^{(*)}_{(s)}+M$, where $M$ stands for a light pseudoscalar or vector meson. We predict that the branching ratio for inclusive non-leptonic two-body weak decays of $J/ψ$ which are realized via the spectator mechanism, can be as large as $1.3 \times 10^{-8}$, in particular, the branching ratio of $J/ψ\to D^{*\pm}_s+ρ^\mp$ can reach $5.3 \times 10^{-9}$. Such values will be marginally accessed by the ability of BESIII which will begin running very soon.

hep-ph↗

arXiv2007

The transition form factors for semi-leptonic weak decays of $J/ψ$ in QCD sum rules

Within the Standard Model, we investigate the semi-leptonic weak decays of $J/ψ$. The various form factors of $J/ψ$ transiting to a single charmed meson ($D^{(*)}_{(d,s)}$) are studied in the framework of the QCD sum rules. These form factors fully determine the rates of the weak semi-leptonic decays of $J/ψ$ and provide valuable information about the non-perturbative QCD effects. Our results indicate that the decay rate of the semi-leptonic weak decay mode $J/ψ\to D^{(*)-}_{s}+e^{+}+ν_{e}$ is at order of $10^{-10}$.

hep-ph↗

arXiv2007

Twist-3 distribution amplitudes of scalar mesons from QCD sum rules

We study the twist-3 distribution amplitudes for scalar mesons made up of two valence quarks based on QCD sum rules. By choosing the proper correlation functions, we derive the moments of the scalar mesons up to the first two order. Making use of these moments, we then calculate the first two Gegenbauer coefficients for twist-3 distribution amplitudes of scalar mesons. It is found that the second Gegenbauer coefficients of scalar density twist-3 distribution amplitudes for $K^{*}_0$ and $f_0$ mesons are quite close to that for $a_0$, which indicates that the SU(3) symmetry breaking effect is tiny here. However, this effect could not be neglected for the forth Gegenbauer coefficients of scalar twist-3 distribution amplitudes between $a_0$ and $f_0$. Besides, we also observe that the first two Gegenbauer coefficients corresponding to the tensor current twist-3 distribution amplitudes for all the $a_0$, $K^{*}_0$ and $f_0$ are very small. The renormalization group evolution of condensates, quark masses, decay constants and moments are considered in our calculations. As a byproduct, it is found that the masses for isospin I=1, ${1 \over 2}$ scalar mesons are around $1.27 \sim 1.41$ GeV and $1.44 \sim 1.56$ GeV respectively, while the mass for isospin state composed of $\bar{s} s$ is $1.62 \sim 1.73$ GeV.

hep-ph↗