arXiv subjects

Daniel Werner

Publications and source records attributed to Daniel Werner.

At least 19 recordsLinked to original sources

arXiv2025

AIRwaves at CheckThat! 2025: Retrieving Scientific Sources for Implicit Claims on Social Media with Dual Encoders and Neural Re-Ranking

Linking implicit scientific claims made on social media to their original publications is crucial for evidence-based fact-checking and scholarly discourse, yet it is hindered by lexical sparsity, very short queries, and domain-specific language. Team AIRwaves ranked second in Subtask 4b of the CLEF-2025 CheckThat! Lab with an evidence-retrieval approach that markedly outperforms the competition baseline. The optimized sparse-retrieval baseline(BM25) achieves MRR@5 = 0.5025 on the gold label blind test set. To surpass this baseline, a two-stage retrieval pipeline is introduced: (i) a first stage that uses a dual encoder based on E5-large, fine-tuned using in-batch and mined hard negatives and enhanced through chunked tokenization and rich document metadata; and (ii) a neural re-ranking stage using a SciBERT cross-encoder. Replacing purely lexical matching with neural representations lifts performance to MRR@5 = 0.6174, and the complete pipeline further improves to MRR@5 = 0.6828. The findings demonstrate that coupling dense retrieval with neural re-rankers delivers a powerful and efficient solution for tweet-to-study matching and provides a practical blueprint for future evidence-retrieval pipelines.

cs.IR↗

arXiv2025

Nonequilibrium transport through an interacting monitored quantum dot

We study the interplay between strong correlations and Markovian dephasing, resulting from monitoring the charge or spin degrees of freedom of a quantum dot described by a dissipative Anderson impurity model. Using the Auxiliary master equation approach we compute the steady-state spectral function and occupation of the dot and discuss the role of dephasing on Kondo physics. Furthermore, we consider a two-lead setup which allows to compute the steady-state current and conductance. We show that the Kondo steady-state is robust to moderate charge dephasing but not to spin dephasing, which we interpret in terms of dephasing-induced heating of low-energy excitations. Finally, we show universal scaling collapse of the non-linear conductance with a dephasing-dependent Kondo scale.

cond-mat.mes-hall↗

arXiv2025

Mixed-configuration approximation for multiorbital systems out of equilibrium

We propose a mixed-configuration approximation based on single-band impurity solvers to efficiently study nonequilibrium multiorbital systems at moderate computational cost. In this work, we merge the approach with the so-called auxiliary master equation approach. As a benchmark, we first show that our approach reproduces the results of quantum Monte Carlo (QMC) for two-orbital impurity models at equilibrium with overall good accuracy, especially for non-degenerate orbitals. We then use our approach as an impurity solver for dynamical mean-field theory (DMFT) to address the case of a realistic two-orbital layered structure, recovering the strong crystal-field-driven charge polarization observed by solving the DMFT self-consistent cycle with QMC, albeit slightly reduced. Finally, we address a prototype nonequilibrium setup by sandwiching this layer between metallic contacts subject to a bias voltage described by different chemical potentials. This simplified model demonstrates our method's potential to access nonequilibrium steady-state behavior in realistic multi-orbital materials. These findings provide a first step toward theoretical studies of the nonequilibrium properties of multi-orbital compounds directly in the real-frequency domain.

cond-mat.str-el↗

arXiv2025

Functional interpolation expansion for nonequilibrium correlated impurities

We present a functional interpolation approach within the auxiliary master equation framework to efficiently and accurately solve correlated impurity problems in nonequilibrium dynamical mean-field theory (DMFT). By leveraging a near-exact auxiliary bath representation, the method estimates corrections via interpolation over a few bath realisations, significantly reducing computational cost and increasing accuracy. We illustrate the approach on the Anderson impurity model and on the Hubbard model within DMFT, capturing equilibrium and long-lived photodoped states.

cond-mat.str-el↗

arXiv2024

Photodriven Mott insulating heterostructures: A steady-state study of impact ionization processes

We investigate the photocurrent and spectral features in a simplified model of a Mott photovoltaic system consisting of a multilayered insulating heterostructure. The central correlated region is coupled to two metallic leads kept at different chemical potentials. A periodic drive applied to the correlated region produces excited doublons and holons across the Mott gap which are then separated by a potential gradient, which mimics the polarization-induced electric field present in oxyde heterostructures. The nonequilibrium Floquet steady-state is addressed by means of dynamical mean-field theory and its Floquet extension, while the so-called auxiliary master equation approach is employed as impurity solver. We find that impact ionization, identified by a kink in the photocurrent as function of the driving frequency, becomes significant and is generally favoured by weak, narrow-band hybridizations to the leads beyond a certain strength of the driving field. On the other hand, in the case of a direct coupling to metallic leads with a flat band, we observe a drastic reduction of impact ionization and of the photocurrent itself.

cond-mat.str-el↗

arXiv2023

Numerically exact simulation of photo-doped Mott insulators

A description of long-lived photo-doped states in Mott insulators is challenging, as it needs to address exponentially separated timescales. We demonstrate how properties of such states can be computed using numerically exact steady state techniques, in particular Quantum Monte Carlo, by using a time-local ansatz for the distribution function with separate Fermi functions for the electron and hole quasiparticles. The simulations show that the Mott gap remains robust to large photo-doping, and the photo-doped state has hole and electron quasiparticles with strongly renormalized properties.

cond-mat.str-el↗

arXiv2023

Impact of disorder and phonons on the Hubbard bands of Mott insulators in strong electric fields

We characterize the current-carrying non-equilibrium steady-state (NESS) in a single-band Hubbard model confronted with a static electric field in the presence of quenched disorder. Beyond linear response regime, the electric field amplitude must be such to compensate for at least half of the band gap in order to have a non-negligible stationary current. As disorder is not expected to dissipate the extra energy injected by the field, optical phonons assisted by a fermionic heat bath serve as dissipation channels for the current-induced Joule heat generated by the accelerated electrons. The NESS of the system is addressed by means of the dynamical mean-field theory using the so-called auxiliary master equation approach as impurity solver. Disorder effects are treated locally via the coherent potential approximation (CPA) and the self-consistent Born (SCB) approach. In the regime in which the two schemes yield similar results, we employ the SCB as it is computationally cheaper than the CPA. We show that, in a purely electronic setup, the disorder-induced dephasing cannot contribute states within the gap but only smear out the edges of the Hubbard bands. When phonons are taken into account, the different nature of disorder-induced dephasing and phonon-related dissipation becomes clear. We show that although both disorder and electron-phonon interaction enhance the current at off-resonant fields, disorder effects play a marginal role since they cannot provide in-gap states which are instead brought about by phonons and represent the privileged relaxation pathway for excited electrons.

cond-mat.str-el↗

arXiv2022

Correlated Mott insulators in a strong electric field: The effects of phonon renormalization

We characterize the response of a Mott insulating system to a static electric field in terms of its conducting and spectral properties. Dissipation is included by a coupling to fermionic baths and to either optical or acoustic phonons. This paper extends and completes the analysis made in a previous work by the authors [arXiv:2207.01921]. In the present work phonons are included diagrammatically within the Migdal approximation by also including self-consistency from the electronic feedback. The nonequilibrium steady-state is addressed by means of the dynamical mean-field theory based on the nonequilibrium Green's function approach, while the so-called auxiliary master equation approach is employed as impurity solver. With optical phonons the self-consistency suppresses the steady-state current at the onset of the metallic phase with respect to the nonself-consistent case. This is due to the interaction of phonons with the hot electrons of the lattice which increases their temperature, thus providing a less effective relaxation channel for the current-induced Joule heat. In addition, in the case of optical phonons the results are essentially independent of the temperature of the fermionic baths, as the latter is sensibly smaller than their characteristic frequency. On the other hand, with acoustic phonons the steady-state current is slightly suppressed by the self-consistent treatment only at field strengths close to half of the gap, away from the metallic phase, and especially at very small phonon frequency. Also, in this case the results seem to slightly depend on the temperature of the fermionic baths.

cond-mat.str-el↗

arXiv2022

On the Chromatic Number of some generalized Kneser Graphs

We determine the chromatic number of the Kneser graph q{\Gamma}_{7,{3,4}} of flags of vectorial type {3, 4} of a rank 7 vector space over the finite field GF(q) for large q and describe the colorings that attain the bound. This result relies heavily, not only on the independence number, but also on the structure of all large independent sets. Furthermore, our proof is more general in the following sense: it provides the chromatic number of the Kneser graphs q{\Gamma}_{2d+1,{d,d+1}} of flags of vectorial type {d, d+1} of a rank 2d+1 vector space over GF(q) for large q as long as the large independent sets of the graphs are only the ones that are known.

math.CO↗

arXiv2020

On the chromatic number of two generalized Kneser graphs

We determine the chromatic number of some graphs of flags in buildings of type $A_4$, namely of the Kneser graphs of flags of type $\{2,4\}$ in the vector spaces $GF(q)^5$ for $q\geq3$, and of the Kneser graph of flags of type $\{2,3\}$ in the vector spaces $GF(q)^5$ for large $q$.

math.CO↗

arXiv2019

Maximal co-cliques in the Kneser graph on plane-solid flags in $PG(6,q)$

For $q>27$ we determine the independence number $\alpha(\Gamma)$ of the Kneser graph $\Gamma$ on plane-solid flags in $PG(6,q)$. More precisely we describe all maximal independent sets of size at least $q^{11}$ and show that every other maximal example has cardinality at most a constant times $q^{10}$.

math.CO↗

arXiv2013

Fixed Parameter Complexity and Approximability of Norm Maximization

The problem of maximizing the $p$-th power of a $p$-norm over a halfspace-presented polytope in $\R^d$ is a convex maximization problem which plays a fundamental role in computational convexity. It has been shown in 1986 that this problem is $\NP$-hard for all values $p \in \mathbb{N}$, if the dimension $d$ of the ambient space is part of the input. In this paper, we use the theory of parametrized complexity to analyze how heavily the hardness of norm maximization relies on the parameter $d$. More precisely, we show that for $p=1$ the problem is fixed parameter tractable but that for all $p \in \mathbb{N} \setminus \{1\}$ norm maximization is W[1]-hard. Concerning approximation algorithms for norm maximization, we show that for fixed accuracy, there is a straightforward approximation algorithm for norm maximization in FPT running time, but there is no FPT approximation algorithm, the running time of which depends polynomially on the accuracy. As with the $\NP$-hardness of norm maximization, the W[1]-hardness immediately carries over to various radius computation tasks in Computational Convexity.

cs.CC↗

arXiv2012

A Lower Bound for Shallow Partitions

Let P be a planar n-point set. A k-partition of P is a subdivision of P into n/k parts of roughly equal size and a sequence of triangles such that each part is contained in a triangle. A line is k-shallow if it has at most k points of P below it. The crossing number of a k-partition is the maximum number of triangles in the partition that any k-shallow line intersects. We give a lower bound of Omega(log (n/k)/loglog(n/k)) for this crossing number, answering a 20-year old question of Matousek.

cs.CG↗

arXiv2011

Erdős-Szekeres and Testing Weak epsilon-Nets are NP-hard in 3 dimensions - and what now?

We consider the computational versions of the Erd\H os-Szekeres theorem and related problems in 3 dimensions. We show that, in constrast to the planar case, no polynomial time algorithm exists for determining the largest (empty) convex subset (unless P=NP) among a set of points, by proving that the corresponding decision problem is NP-hard. This answers a question by Dobkin, Edelsbrunner and Overmars from 1990. As a corollary, we derive a similar result for the closely related problem of testing weak epsilon-nets in R^3. Answering a question by Chazelle et al. from 1995, our reduction shows that the problem is co-NP-hard. This is work in progress - we are still trying to find a smart approximation algorithm for the problems.

cs.CG↗

arXiv2011

Approximating Tverberg Points in Linear Time for Any Fixed Dimension

Let P be a d-dimensional n-point set. A Tverberg-partition of P is a partition of P into r sets P_1, ..., P_r such that the convex hulls conv(P_1), ..., conv(P_r) have non-empty intersection. A point in the intersection of the conv(P_i)'s is called a Tverberg point of depth r for P. A classic result by Tverberg implies that there always exists a Tverberg partition of size n/(d+1), but it is not known how to find such a partition in polynomial time. Therefore, approximate solutions are of interest. We describe a deterministic algorithm that finds a Tverberg partition of size n/4(d+1)^3 in time d^{O(log d)} n. This means that for every fixed dimension we can compute an approximate Tverberg point (and hence also an approximate centerpoint) in linear time. Our algorithm is obtained by combining a novel lifting approach with a recent result by Miller and Sheehy (2010).

cs.CG↗

arXiv2011

Hardness of discrepancy computation and epsilon-net verification in high dimension

Discrepancy measures how uniformly distributed a point set is with respect to a given set of ranges. There are two notions of discrepancy, namely continuous discrepancy and combinatorial discrepancy. Depending on the ranges, several possible variants arise, for example star discrepancy, box discrepancy, and discrepancy of half-spaces. In this paper, we investigate the hardness of these problems with respect to the dimension d of the underlying space. All these problems are solvable in time {n^O(d)}, but such a time dependency quickly becomes intractable for high-dimensional data. Thus it is interesting to ask whether the dependency on d can be moderated. We answer this question negatively by proving that the canonical decision problems are W[1]-hard with respect to the dimension. This is done via a parameterized reduction from the Clique problem. As the parameter stays linear in the input parameter, the results moreover imply that these problems require {n^Ω(d)} time, unless 3-Sat can be solved in {2^o(n)} time. Further, we derive that testing whether a given set is an ε-net with respect to half-spaces takes {n^Ω(d)} time under the same assumption. As intermediate results, we discover the W[1]-hardness of other well known problems, such as determining the largest empty star inside the unit cube. For this, we show that it is even hard to approximate within a factor of {2^n}.

cs.CG↗

arXiv2010

Polynomial Bounds on the Slicing Number

NOTE: Unfortunately, most of the results mentioned here were already known under the name of "d-separated interval piercing". The result that T_d(m) exists was first proved by Gyaŕfaś and Lehel in 1970, see [5]. Later, the result was strengthened by Kaŕolyi and Tardos [9] to match our result. Moreover, their proof (in a different notation, of course) uses ideas very similar to ours and leads to a similar recurrence. Also, our conjecture turns out to be right and was proved for the 2-dimensional case by Tardos and for the general case by Kaiser [8]. An excellent survey article ("Transversals of d-intervals') is available on http://www.renyi.hu/~tardos. Still, we leave this paper available to the public on http://page.mi.fu-berlin.de/dawerner, also because one might find the references useful. ----- We study the following Gallai-type of problem: Assume that we are given a family X of convex objects in R^d such that among any subset of size m, there is an axis-parallel hyperplane intersecting at least two of the objects. What can we say about the number of axis-parallel hyperplanes that sufficient to intersect all sets in the family? In this paper, we show that this number T_d(m) exists, i.e., depends only on m and the dimension d, but not on the size of the set X. First, we derive a very weak super-exponential bound. Using this result, by a simple proof we are able to show that this number is even polynomially bounded for any fixed d. We partly answer open problem 74 on http://maven.smith.edu/~orourke/TOPP/, where the planar case is considered, by improving the best known exponential bound to O(m^2).

cs.CG↗

arXiv2009

Fixed-parameter tractability and lower bounds for stabbing problems

We study the following general stabbing problem from a parameterized complexity point of view: Given a set $\mathcal S$ of $n$ translates of an object in $\Rd$, find a set of $k$ lines with the property that every object in $\mathcal S$ is ''stabbed'' (intersected) by at least one line. We show that when $S$ consists of axis-parallel unit squares in $\Rtwo$ the (decision) problem of stabbing $S$ with axis-parallel lines is W[1]-hard with respect to $k$ (and thus, not fixed-parameter tractable unless FPT=W[1]) while it becomes fixed-parameter tractable when the squares are disjoint. We also show that the problem of stabbing a set of disjoint unit squares in $\Rtwo$ with lines of arbitrary directions is W[1]--hard with respect to $k$. Several generalizations to other types of objects and lines with arbitrary directions are also presented. Finally, we show that deciding whether a set of unit balls in $\Rd$ can be stabbed by one line is W[1]--hard with respect to the dimension $d$.

cs.CG↗