Source author record

Lei Fu

Lei Fu appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

math.AG astro-ph.HE math.NT Artificial Intelligence Computation and Language

Catalog footprint

What is connected

11works

5topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2025arXiv

Companion Agents: A Table-Information Mining Paradigm for Text-to-SQL

Large-scale Text-to-SQL benchmarks such as BIRD typically assume complete and accurate database annotations as well as readily available external knowledge, which fails to reflect common industrial settings where annotations are missing, incomplete, or erroneous. This mismatch substantially limits the real-world applicability of state-of-the-art (SOTA) Text-to-SQL systems. To bridge this gap, we explore a database-centric approach that leverages intrinsic, fine-grained information residing in relational databases to construct missing evidence and improve Text-to-SQL accuracy under annotation-scarce conditions. Our key hypothesis is that when a query requires multi-step reasoning over extensive table information, existing methods often struggle to reliably identify and utilize the truly relevant knowledge. We therefore propose to "cache" query-relevant knowledge on the database side in advance, so that it can be selectively activated at inference time. Based on this idea, we introduce Companion Agents (CA), a new Text-to-SQL paradigm that incorporates a group of agents accompanying database schemas to proactively mine and consolidate hidden inter-table relations, value-domain distributions, statistical regularities, and latent semantic cues before query generation. Experiments on BIRD under the fully missing evidence setting show that CA recovers +4.49 / +4.37 / +14.13 execution accuracy points on RSL-SQL / CHESS / DAIL-SQL, respectively, with larger gains on the Challenging subset +9.65 / +7.58 / +16.71. These improvements stem from CA's automatic database-side mining and evidence construction, suggesting a practical path toward industrial-grade Text-to-SQL deployment without reliance on human-curated evidence.

preprint2022arXiv

Gravitational wave signals in the deci-Hz range from neutrinos during the proto-neutron star cooling phase

We investigate the gravitational waves (GWs) at low frequencies produced by neutrinos that are emitted anisotropically from the proto-neutron star (PNS) during its cooling phase that lasts about a minute. We first give a formulation based on the spherical-harmonic expansion of the neutrino luminosity to obtain the gravitational waveform as well as the characteristic strain. In the absence of multi-dimensional simulations of PNS cooling, from which we can extract reliable data on the neutrino luminosities as a function of solid angle, we construct them by hand. In the first model, the time evolution is approximated by piece-wise exponential functions (PEFs); in the second model we employ the time evolution obtained in a 1D cooling simulation for all harmonic components for simplicity; In both cases, we consider not only axisymmetric components but also non-axisymmetric ones; for the latter, in particular, we consider as the third model axisymetric neutrino emissions that are misaligned with the rotation axis and, as a result, rotate with the PNS. We find from the first model that the decay times in PEF at late phases can be inferred from the positions of bumps and dips in the characteristic strain of the GW in the case of a slow cooling, whereas it may be obtained by identifying the positions of slope change in the case of rapid cooling, which may be induced by convections in PNS. We also find from the first model that the rotation of PNS can be reflected in GW waveforms. Finally, we compared the characteristic strains with sensitivity curves GW detectors, finding that above phenomena are expected to be detected by DECIGO if the PNS is located at 10kpc.

preprint2020arXiv

On Katz's $(A,B)$-exponential sums

We deduce Katz's theorems for $(A,B)$-exponential sums over finite fields using $\ell$-adic cohomology and a theorem of Denef-Loeser, removing the hypothesis that $A+B$ is relatively prime to the characteristic $p$. In some degenerate cases, the Betti number estimate is improved using toric decomposition and Adolphson-Sperber's bound for the degree of $L$-functions. Applying the facial decomposition theorem in \cite{W1}, we prove that the universal family of $(A,B)$-polynomials is generically ordinary for its $L$-function when $p$ is in certain arithmetic progression.

preprint2020arXiv

Unirational Differential Curves and Differential Rational Parametrizations

In this paper, we study unirational differential curves and the corresponding differential rational parametrizations. We first investigate basic properties of proper differential rational parametrizations for unirational differential curves. Then we show that the implicitization problem of proper linear differential rational parametric equations can be solved by means of differential resultants. Furthermore, for linear differential curves, we give an algorithm to determine whether an implicitly given linear differential curve is unirational and, in the affirmative case, to compute a proper differential rational parametrization for the differential curve.

preprint2016arXiv

$\ell$-adic GKZ hypergeometric sheaf and exponential sums

To a torus action on a complex vector space, Gelfand, Kapranov and Zelevinsky introduce a system of differential equations, called the GKZ hypergeometric system. Its solutions are GKZ hypergeometric functions. We study the $\ell$-adic counterpart of the GKZ hypergeometric system, which we call the $\ell$-adic GKZ hypergeometric sheaf. It is an object in the derived category of $\ell$-adic sheaves on the affine space over a finite field. Traces of Frobenius on stalks of this object at rational points of the affine space define the hypergeometric functions over the finite field introduced by Gelfand and Graev. We prove that the $\ell$-adic GKZ hypergeometric sheaf is perverse, calculate its rank, and prove that it is irreducible under the non-resonance condition. We also study the weight filtration of the GKZ hypergeometric sheaf, determine its lisse locus, and apply our result to the study of weights of twisted exponential sums.

preprint2013arXiv

A Thom-Sebastiani Theorem in Characteristic p

Let $k$ be a perfect field of characteristic $p$, let $f_i:X_i\to\mathbb A_k^1$ $(i=1,2)$ be two $k$-morphism of finite type, and let $f:X_1\times_k X_2\to \mathbb A_k^1$ be the morphism defined by $f(z_1,z_2)=f_1(z_1)+f_2(z_2)$. For each $i\in\{1,2\}$, let $x_i$ be a $k$-rational point in the fiber $f_i^{-1}(0)$ such that $f_i$ is smooth on $X_i-\{x_i\}$. Using the $\ell$-adic Fourier transformation and the stationary phase principle of Laumon, we prove that the vanishing cycle of $f$ at $x=(x_1,x_2)$ is the convolution product of the vanishing cycles of $f_i$ at $x_i$ $(i=1,2)$.

preprint2013arXiv

Population synthesis of young isolated neutron stars: the effect of fallback disk accretion and magnetic field evolution

The spin evolution of isolated neutron stars (NSs) is dominatd by their magnetic fields. The measured braking indices of young NSs show that the spin-down mechanism due to magnetic dipole radiation with constant magnetic fields is inadequate. Assuming that the NS magnetic field is buried by supernova fallback matter and re-emerges after accretion stops, we carry out Monte-Carlo simulation of the evolution of young NSs, and show that most of the pulsars have the braking indices ranging from -1 to 3. The results are compatible with the observational data of NSs associated with supernova remnants. They also suggest that the initial spin periods of NSs might occupy a relatively wide range.

preprint2012arXiv

Could SXP 1062 be an Accreting Magnetar?

In this work we explore the possible evolutionary track of the neutron star in the newly discovered Be/X-ray binary SXP 1062, which is believed to be the first X-ray pulsar associated with a supernova remnant. Although no cyclotron feature has been detected to indicate the strength of the neutron star's magnetic field, we show that it may be $\ga 10^{14} $G. If so SXP 1062 may belong to the accreting magnetars in binary systems. We attempt to reconcile the short age and long spin period of the pulsar taking account of different initial parameters and spin-down mechanisms of the neutron star. Our calculated results show that, to spin down to a period $\sim 1000 $s within $10-40 $kyr requires efficient propeller mechanisms. In particular, the model for angular momentum loss under energy conservation seems to be ruled out.

preprint2012arXiv

Deformation of $\ell$-adic sheaves with Undeformed Local Monodromy

Let $X$ be a smooth connected algebraic curve over an algebraically closed field $k$. We study the deformation of $\ell$-adic Galois representations of the function field of $X$ while keeping the local Galois representations at all places undeformed.

preprint2011arXiv

On Pulsar-Driven Mass Ejection in Low-Mass X-ray Binaries

There is accumulating evidence for mass ejection in low-mass X-ray binaries (LMXBs) driven by radio pulsar activity during X-ray quiescence. In this paper we consider the condition for mass ejection by comparing the radiation pressure from a millisecond pulsar, and the gas pressure at the inner Lagrange point or at the surrounding accretion disk. We calculate the critical spin period of the pulsar below which mass ejection is allowed. Combining with the evolution of the mass transfer rate, we present constraints on the orbital periods of the systems. We show that mass ejection could happen in both wide and compact LMXBs. It may be caused by transient accretion due to thermal instability in the accretion disks in the former, and irradiation-driven mass-transfer cycles in the latter.

preprint2010arXiv

A Tauberian Theorem for $\ell$-adic Sheaves on $\mathbb A^1$

Let $K\in L^1(\mathbb R)$ and let $f\in L^\infty(\mathbb R)$ be two functions on $\mathbb R$. The convolution $$(K\ast f)(x)=\int_{\mathbb R}K(x-y)f(y)dy$$ can be considered as an average of $f$ with weight defined by $K$. Wiener's Tauberian theorem says that under suitable conditions, if $$\lim_{x\to \infty}(K\ast f)(x)=\lim_{x\to \infty} (K\ast A)(x)$$ for some constant $A$, then $$\lim_{x\to \infty}f(x)=A.$$ We prove the following $\ell$-adic analogue of this theorem: Suppose $K,F, G$ are perverse $\ell$-adic sheaves on the affine line $\mathbb A$ over an algebraically closed field of characteristic $p$ ($p\not=\ell$). Under suitable conditions, if $$(K\ast F)|_{η_\infty}\cong (K\ast G)|_{η_\infty},$$ then $$F|_{η_\infty}\cong G|_{η_\infty},$$ where $η_\infty$ is the spectrum of the local field of $\mathbb A$ at $\infty$.