Digital Humanities and Computational Literary Studies apply text mining methods to investigate literature. Such automated approaches enable quantitative studies on large corpora which would not be feasible by manual inspection alone. However, due to copyright restrictions, the availability of relevant digitized literary works is limited. Derived Text Formats (DTFs) have been proposed as a solution. Here, textual materials are transformed in such a way that copyright-critical features are removed, but that the use of certain analytical methods remains possible. Contextualized word embeddings produced by transformer-encoders (like BERT) are promising candidates for DTFs because they allow for state-of-the-art performance on various analytical tasks and, at first sight, do not disclose the original text. However, in this paper we demonstrate that under certain conditions the reconstruction of the original copyrighted text becomes feasible and its publication in the form of contextualized token representations is not safe. Our attempts to invert BERT suggest, that publishing the encoder as a black box together with the contextualized embeddings is critical, since it allows to generate data to train a decoder with a reconstruction accuracy sufficient to violate copyright laws.

Large language models have become the latest trend in natural language processing, heavily featuring in the digital tools we use every day. However, their replies often reflect a narrow cultural viewpoint that overlooks the diversity of global users. This missing capability could be referred to as cultural reasoning, which we define here as the capacity of a model to recognise culture-specific knowledge values and social norms, and to adjust its output so that it aligns with the expectations of individual users. Because culture shapes interpretation, emotional resonance, and acceptable behaviour, cultural reasoning is essential for identity-aware AI. When this capacity is limited or absent, models can sustain stereotypes, ignore minority perspectives, erode trust, and perpetuate hate. Recent empirical studies strongly suggest that current models default to Western norms when judging moral dilemmas, interpreting idioms, or offering advice, and that fine-tuning on survey data only partly reduces this tendency. The present evaluation methods mainly report static accuracy scores and thus fail to capture adaptive reasoning in context. Although broader datasets can help, they cannot alone ensure genuine cultural competence. Therefore, we argue that cultural reasoning must be treated as a foundational capability alongside factual accuracy and linguistic coherence. By clarifying the concept and outlining initial directions for its assessment, a foundation is laid for future systems to be able to respond with greater sensitivity to the complex fabric of human culture.

The ability of Large Language Models (LLMs) to mimic human behavior triggered a plethora of computational social science research, assuming that empirical studies of humans can be conducted with AI agents instead. Since there have been conflicting research findings on whether and when this hypothesis holds, there is a need to better understand the differences in their experimental designs. We focus on replicating the behavior of social network users with the use of LLMs for the analysis of communication on social networks. First, we provide a formal framework for the simulation of social networks, before focusing on the sub-task of imitating user communication. We empirically test different approaches to imitate user behavior on X in English and German. Our findings suggest that social simulations should be validated by their empirical realism measured in the setting in which the simulation components were fitted. With this paper, we argue for more rigor when applying generative-agent-based modeling for social simulation.

This paper targets the automated extraction of components of argumentative information and their relations from natural language text. Moreover, we address a current lack of systems to provide complete argumentative structure from arbitrary natural language text for general usage. We present an argument mining pipeline as a universally applicable approach for transforming German and English language texts to graph-based argument representations. We also introduce new methods for evaluating the results based on existing benchmark argument structures. Our results show that the generated argument graphs can be beneficial to detect new connections between different statements of an argumentative text. Our pipeline implementation is publicly available on GitHub.

Are AI systems truly representing human values, or merely averaging across them? Our study suggests a concerning reality: Large Language Models (LLMs) fail to represent diverse cultural moral frameworks despite their linguistic capabilities. We expose significant gaps between AI-generated and human moral intuitions by applying the Moral Foundations Questionnaire across 19 cultural contexts. Comparing multiple state-of-the-art LLMs' origins against human baseline data, we find these models systematically homogenize moral diversity. Surprisingly, increased model size doesn't consistently improve cultural representation fidelity. Our findings challenge the growing use of LLMs as synthetic populations in social science research and highlight a fundamental limitation in current AI alignment approaches. Without data-driven alignment beyond prompting, these systems cannot capture the nuanced, culturally-specific moral intuitions. Our results call for more grounded alignment objectives and evaluation metrics to ensure AI systems represent diverse human values rather than flattening the moral landscape.

Aligning machine learning systems with human expectations is mostly attempted by training with manually vetted human behavioral samples, typically explicit feedback. This is done on a population level since the context that is capturing the subjective Point-Of-View (POV) of a concrete person in a specific situational context is not retained in the data. However, we argue that alignment on an individual level can boost the subjective predictive performance for the individual user interacting with the system considerably. Since perception differs for each person, the same situation is observed differently. Consequently, the basis for decision making and the subsequent reasoning processes and observable reactions differ. We hypothesize that individual perception patterns can be used for improving the alignment on an individual level. We test this, by integrating perception information into machine learning systems and measuring their predictive performance wrt.~individual subjective assessments. For our empirical study, we collect a novel data set of multimodal stimuli and corresponding eye tracking sequences for the novel task of Perception-Guided Crossmodal Entailment and tackle it with our Perception-Guided Multimodal Transformer. Our findings suggest that exploiting individual perception signals for the machine learning of subjective human assessments provides a valuable cue for individual alignment. It does not only improve the overall predictive performance from the point-of-view of the individual user but might also contribute to steering AI systems towards every person's individual expectations and values.

Scientific writing builds upon already published papers. Manual identification of publications to read, cite or consider as related papers relies on a researcher's ability to identify fitting keywords or initial papers from which a literature search can be started. The rapidly increasing amount of papers has called for automatic measures to find the desired relevant publications, so-called paper recommendation systems. As the number of publications increases so does the amount of paper recommendation systems. Former literature reviews focused on discussing the general landscape of approaches throughout the years and highlight the main directions. We refrain from this perspective, instead we only consider a comparatively small time frame but analyse it fully. In this literature review we discuss used methods, datasets, evaluations and open challenges encountered in all works first released between January 2019 and October 2021. The goal of this survey is to provide a comprehensive and complete overview of current paper recommendation systems.

Representing relevant information of a traffic scene and understanding its environment is crucial for the success of autonomous driving. Modeling the surrounding of an autonomous car using semantic relations, i.e., how different traffic participants relate in the context of traffic rule based behaviors, is hardly been considered in previous work. This stems from the fact that these relations are hard to extract from real-world traffic scenes. In this work, we model traffic scenes in a form of spatial semantic scene graphs for various different predictions about the traffic participants, e.g., acceleration and deceleration. Our learning and inference approach uses Graph Neural Networks (GNNs) and shows that incorporating explicit information about the spatial semantic relations between traffic participants improves the predicdtion results. Specifically, the acceleration prediction of traffic participants is improved by up to 12% compared to the baselines, which do not exploit this explicit information. Furthermore, by including additional information about previous scenes, we achieve 73% improvements.

Extended exposure to virtual reality environments can induce motion sickness, often referred to as cybersickness, which may lead to physiological stress responses and impaired cognitive performance. This study investigates the aftereffects of VR-induced motion sickness with a focus on physiological stress markers and working memory performance. Using a carousel simulation to elicit cybersickness, we assessed subjective discomfort (SSQ, FMS), physiological stress (salivary cortisol, alpha-amylase, electrodermal activity, heart rate), and cognitive performance (n-Back task) over a 90-minute post-exposure period. Our findings demonstrate a significant increase in both subjective and physiological stress indicators following VR exposure, accompanied by a decline in working memory performance. Notably, delayed symptom progression was observed in a substantial proportion of participants, with some reporting peak symptoms up to 90 minutes post-stimulation. Salivary cortisol levels remained elevated throughout the observation period, indicating prolonged stress recovery. These results highlight the need for longer washout phases in XR research and raise safety concerns for professional applications involving post-exposure task performance.

We say that a (multi)graph $G = (V,E)$ has geometric thickness $t$ if there exists a straight-line drawing $\varphi : V \rightarrow \mathbb{R}^2$ and a $t$-coloring of its edges where no two edges sharing a point in their relative interior have the same color. The \textsc{Geometric Thickness} problem asks whether a given multigraph has geometric thickness at most $t$. This problem was shown to be NP-hard for $t=2$ [Durocher, Gethner, and Mondal, CG 2016]. In this paper, we settle the computational complexity of \textsc{Geometric Thickness} by showing that it is $\exists \mathbb{R}$-complete already for thickness $30$. Moreover, our reduction shows that the problem is $\exists \mathbb{R}$-complete for$4392$-planar graphs, where a graph is$k$-planar if it admits a topological drawing with at most $k$ crossings per edge. In the course of our paper, we answer previous questions on geometric thickness and on other related problems, in particular that simultaneous graph embeddings of $31$ edge-disjoint graphs and pseudo-segment stretchability with chromatic number $30$ are $\exists \mathbb{R}$-complete.

We provide a correction to the sufficient conditions under which closed-form expressions for the optimal Lagrange multiplier are provided in arXiv:2112.13138 [math.OC]. We first present a simple counterexample where the original conditions are insufficient, highlight where the original proof fails, and then provide modified conditions along with a correct proof of their validity. Finally, although the original paper discusses modifications to their method for problems that may not satisfy any sufficient conditions, we substantiate that discussion along two directions. We first show that computing an optimal Lagrange multiplier can still be done in polynomial time. We then provide complete and correct versions of the corresponding Benders and column-and-constraint generation algorithms in which the original method is used. We also discuss the implications of our findings on computational performance.

The paper motivates high dimensional smoothing with penalized splines and its numerical calculation in an efficient way. If smoothing is carried out over three or more covariates the classical tensor product spline bases explode in their dimension bringing the estimation to its numerical limits. A recent approach by Siebenborn and Wagner(2019) circumvents storage expensive implementations by proposing matrix-free calculations which allows to smooth over several covariates. We extend their approach here by linking penalized smoothing and its Bayesian formulation as mixed model which provides a matrix-free calculation of the smoothing parameter to avoid the use of high-computational cross validation. Further, we show how to extend the ideas towards generalized regression models. The extended approach is applied to remote sensing satellite data in combination with spatial smoothing.

This paper provides a first contribution to port-Hamiltonian modeling of district heating networks. By introducing a model hierarchy of flow equations on the network, this work aims at a thermodynamically consistent port-Hamiltonian embedding of the partial differential-algebraic systems. We show that a spatially discretized network model describing the advection of the internal energy density with respect to an underlying incompressible stationary Euler-type hydrodynamics can be considered as a parameter-dependent finite-dimensional port-Hamiltonian system. Moreover, we present an infinite-dimensional port-Hamiltonian formulation for a compressible instationary thermodynamic fluid flow in a pipe. Based on these first promising results, we raise open questions and point out research perspectives concerning structure-preserving discretization, model reduction, and optimization.

Current deep learning methods for object recognition are purely data-driven and require a large number of training samples to achieve good results. Due to their sole dependence on image data, these methods tend to fail when confronted with new environments where even small deviations occur. Human perception, however, has proven to be significantly more robust to such distribution shifts. It is assumed that their ability to deal with unknown scenarios is based on extensive incorporation of contextual knowledge. Context can be based either on object co-occurrences in a scene or on memory of experience. In accordance with the human visual cortex which uses context to form different object representations for a seen image, we propose an approach that enhances deep learning methods by using external contextual knowledge encoded in a knowledge graph. Therefore, we extract different contextual views from a generic knowledge graph, transform the views into vector space and infuse it into a DNN. We conduct a series of experiments to investigate the impact of different contextual views on the learned object representations for the same image dataset. The experimental results provide evidence that the contextual views influence the image representations in the DNN differently and therefore lead to different predictions for the same images. We also show that context helps to strengthen the robustness of object recognition models for out-of-distribution images, usually occurring in transfer learning tasks or real-world scenarios.

In this work we propose a general nonmonotone line-search method for nonconvex multi\-objective optimization problems with convex constraints. At the $k$th iteration, the degree of nonmonotonicity is controlled by a vector $\nu_{k}$ with nonnegative components. Different choices for $\nu_{k}$ lead to different nonmonotone step-size rules. Assuming that the sequence $\left\{\nu_{k}\right\}_{k\geq 0}$ is summable, and that the $i$th objective function has Hölder continuous gradient with smoothness parameter $\theta_i \in(0,1]$, we show that the proposed method takes no more than $\mathcal{O}\left(\epsilon^{-\left(1+\frac{1}{\theta_{\min}}\right)}\right)$ iterations to find a $\epsilon$-approximate Pareto critical point for a problem with $m$ objectives and $\theta_{\min}= \min_{i=1,\dots, m} \{\theta_i\}$. In particular, this complexity bound applies to the methods proposed by Drummond and Iusem (Comput. Optim. Appl. 28: 5--29, 2004), by Fazzio and Schuverdt (Optim. Lett. 13: 1365--1379, 2019), and by Mita, Fukuda and Yamashita (J. Glob. Optim. 75: 63--90, 2019). The generality of our approach also allows the development of new methods for multiobjective optimization. As an example, we propose a new nonmonotone step-size rule inspired by the Metropolis criterion. Preliminary numerical results illustrate the benefit of nonmonotone line searches and suggest that our new rule is particularly suitable for multiobjective problems in which at least one of the objectives has many non-global local minimizers.

Contemporary web pages with increasingly sophisticated interfaces rival traditional desktop applications for interface complexity and are often called web applications or RIA (Rich Internet Applications). They often require the execution of JavaScript in a web browser and can call AJAX requests to dynamically generate the content, reacting to user interaction. From the automatic data acquisition point of view, thus, it is essential to be able to correctly render web pages and mimic user actions to obtain relevant data from the web page content. Briefly, to obtain data through existing Web interfaces and transform it into structured form, contemporary wrappers should be able to: 1) interact with sophisticated interfaces of web applications; 2) precisely acquire relevant data; 3) scale with the number of crawled web pages or states of web application; 4) have an embeddable programming API for integration with existing web technologies. OXPath is a state-of-the-art technology, which is compliant with these requirements and demonstrated its efficiency in comprehensive experiments. OXPath integrates Firefox for correct rendering of web pages and extends XPath 1.0 for the DOM node selection, interaction, and extraction. It provides means for converting extracted data into different formats, such as XML, JSON, CSV, and saving data into relational databases. This tutorial explains main features of the OXPath language and the setup of a suitable working environment. The guidelines for using OXPath are provided in the form of prototypical examples.

We analyze an irreversible investment decision for a project which yields a flow of future operating profits given by a geometric Brownian motion with unknown drift. In contrast to similar optimal stopping problems with incomplete information, the agent's payoff now depends directly on the unknown drift and not only indirectly through the underlying dynamics. Hence, many standard arguments are not applicable. Nonetheless, we show that it is optimal to invest in the project if the current profit level exceeds a threshold depending on the current belief for the true state of the unknown drift. These thresholds are described by a boundary function, for which we establish structural properties like monotonicity and continuity. To prove these, we identify a central class of stopping times with useful features. Moreover, we characterize the boundary function as the unique solution of a nonlinear integral equation. Building on this characterization we compute the boundary function numerically and investigate the value of information.

We investigate saturated geometric drawings of graphs with geometric thickness $k$, where no edge can be added without increasing $k$. We establish lower and upper bounds on the number of edges in such drawings if the vertices lie in convex position. We also study the more restricted version where edges are precolored, and for $k=2$ the case for vertices in non-convex position.

A dichotomous ordinal graph consists of an undirected graph with a partition of the edges into short and long edges. A geometric realization of a dichotomous ordinal graph $G$ in a metric space $X$ is a drawing of $G$ in $X$ in which every long edge is strictly longer than every short edge. We call a graph $G$ pandichotomous in $X$ if $G$ admits a geometric realization in $X$ for every partition of its edge set into short and long edges. We exhibit a very close relationship between the degeneracy of a graph $G$ and its pandichotomic Euclidean or spherical dimension, that is, the smallest dimension $k$ such that $G$ is pandichotomous in $\mathbb{R}^k$ or the sphere $\mathbb{S}^k$, respectively. First, every $d$-degenerate graph is pandichotomous in $\mathbb{R}^{d}$ and $\mathbb{S}^{d-1}$ and these bounds are tight for the sphere and for $\mathbb{R}^2$ and almost tight for $\mathbb{R}^d$, for $d\ge 3$. Second, every $n$-vertex graph that is pandichotomous in $\mathbb{R}^k$ has at most $\mu kn$ edges, for some absolute constant \mu<7.23. This shows that the pandichotomic Euclidean dimension of any graph is linearly tied to its degeneracy and in the special cases $k\in \{1,2\}$ resolves open problems posed by Alam, Kobourov, Pupyrev, and Toeniskoetter. Further, we characterize which complete bipartite graphs are pandichotomous in $\mathbb{R}^2$: These are exactly the $K_{m,n}$ with $m\le 3$ or $m=4$ and $n\le 6$. For general bipartite graphs, we can guarantee realizations in $\mathbb{R}^2$ if the short or the long subgraph is constrained: namely if the short subgraph is outerplanar or a subgraph of a rectangular grid, or if the long subgraph forms a caterpillar.