arXiv subjects

Verena Wolf

Publications and source records attributed to Verena Wolf.

At least 19 recordsLinked to original sources

arXiv2026

Modularizing Educational LLM-Agency for Fostering Responsible Learning Assistance

The widespread adoption of AI chatbots in education will drastically change learning, making responsible deployment a critical concern. While large language models (LLMs) might have access to sources discussing insights from educational sciences, they are not particularly inclined to adhere to pedagogical concepts, risking negative effects on the learning process, such as a loss of transfer capabilities, critical thinking, or creativity. In this paper, we introduce an agentic AI chatbot architecture assisting students with exercise solving, specifically designed to contribute to more responsible AI use in education. We base our conceptual development on the identification of several desiderata for responsible LLM-based educational systems, argue for the structural shortcomings inherent in monolithic, out-of-the-box solutions, and instead suggest modularizing the agentic architecture. We propose specific modules for different stages of exercise solving, enabling incorporation of targeted pedagogical advice, guiding students through the learning process in a more controllable, transparent, and overseeable manner.

cs.AI↗

arXiv2026

B-cos GNNs: Faithful Explanations through Dynamic Linearity

We introduce B-cos GNNs, an inherently explainable class of graph neural networks whose predictions decompose exactly into per-node, per-feature contributions via a single input-dependent linear map. B-cos GNNs use linear (sum-based) aggregation and replace non-linear message and update functions with B-cos transforms. This induces meaningful, task-specific weight-input alignment that is directly accessible through the model's dynamic linearity. Instance-level explanations follow from a single forward and backward pass, requiring no auxiliary explainer, modified learning objective, or perturbation procedure. Instantiated as a GIN, our approach trades small losses in predictive accuracy for state-of-the-art explainability across diverse synthetic and real-world benchmarks, producing explanations orders of magnitude faster than post-hoc baselines.

cs.LG↗

arXiv2026

From Fragments to Flares: Migration, Tidal Disruption, and Observable Bursts in Massive Protostellar Disks

We investigate how resolving the inner few astronomical units of a massive protostellar disk affects the migration, disruption, and accretion signatures of an inward-moving fragment. In particular, we aim to determine whether the predicted burst strength and duration depend on the adopted sink cell size. We present a new three-dimensional radiation-hydrodynamic simulation of a $\sim$5$M_{\odot}$ protostar surrounded by a self-gravitating disk, comparing the original 30 AU sink model to a refined model with a 1 AU sink that resolves the inner disk. The resulting gas structures are post-processed with radiative transfer calculations to derive synthetic photometry and multi-band images. Both simulations produce a major accretion burst as a migrating fragment is tidally disrupted, but their detailed behavior differs markedly. The refined model shows faster migration, a complete tidal disruption of the fragment, and a shorter, sharper outburst (more consistent with observations) with nearly the same peak accretion rate as the 30 AU model, which yields a broader, smoother event. The refined run produces much stronger near- and mid-infrared emission, reflecting the formation of a compact, hot inner disk. Resolving the inner few AU qualitatively changes the dynamics and observable appearance of fragment-driven bursts. Diffuse fragment disruption can reproduce decade-long events, but the much shorter ($<$3 yr) bursts observed in some massive protostars likely require the tidal disruption of more compact objects such as second Larson cores. Our trajectory analysis indicates that second Larson cores can migrate sufficiently close to the star to be tidally destroyed, offering a plausible mechanism for the fastest FU-Ori-like bursts observed in massive protostars.

astro-ph.SR↗

arXiv2025

Sub-millimeter wavelength protostellar accretion rate monitoring with AtLAST

How a star forms is a fundamental question in astrophysics. In the earliest stages of protostellar evolution high extinction prevents a direct study of the accretion processes and their temporal evolution. Monitoring the variations of the accretion luminosity in a large protostar sample over decades is needed to reveal how protostars accrete -- in major bursts or in a quasi-steady fashion. We here argue that a large ground based sub-millimeter single-dish facility with a wide FoV is required to fulfill this task.

astro-ph.IM↗

arXiv2025

Insights gained from the Light Echo of Cepheus A HW2

Cepheus A HW2 is a well-studied massive young stellar object (MYSO) featuring Class II methanol masers. Recently, certain maser components have been found to flare every five years. This period went undetected in (NEO)WISE photometry due to detector saturation. However, difference imaging revealed a light echo (LE), representing a variability record. Its periodicity is similar to that of the masers. Thereby, the mid-IR light curve since 2007 could be reconstructed. Phase and period maps provide information on the circumstellar dust distribution and viewing geometry. This is the first time an LE has been used for reverberation mapping of a YSO environment and expanding its light curve into the past.

astro-ph.SR↗

arXiv2025

Per-Domain Generalizing Policies: On Validation Instances and Scaling Behavior

Recent work has shown that successful per-domain generalizing action policies can be learned. Scaling behavior, from small training instances to large test instances, is the key objective; and the use of validation instances larger than training instances is one key to achieve it. Prior work has used fixed validation sets. Here, we introduce a method generating the validation set dynamically, on the fly, increasing instance size so long as informative and feasible.We also introduce refined methodology for evaluating scaling behavior, generating test instances systematically to guarantee a given confidence in coverage performance for each instance size. In experiments, dynamic validation improves scaling behavior of GNN policies in all 9 domains used.

cs.LG↗

arXiv2025

Exploring Molecule Generation Using Latent Space Graph Diffusion

Generating molecular graphs is a challenging task due to their discrete nature and the competitive objectives involved. Diffusion models have emerged as SOTA approaches in data generation across various modalities. For molecular graphs, graph neural networks (GNNs) as a diffusion backbone have achieved impressive results. Latent space diffusion, where diffusion occurs in a low-dimensional space via an autoencoder, has demonstrated computational efficiency. However, the literature on latent space diffusion for molecular graphs is scarce, and no commonly accepted best practices exist. In this work, we explore different approaches and hyperparameters, contrasting generative flow models (denoising diffusion, flow matching, heat dissipation) and architectures (GNNs and E(3)-equivariant GNNs). Our experiments reveal a high sensitivity to the choice of approach and design decisions. Code is made available at github.com/Prashanth-Pombala/Molecule-Generation-using-Latent-Space-Graph-Diffusion.

cs.LG↗

arXiv2024

Acting for the Right Reasons: Creating Reason-Sensitive Artificial Moral Agents

We propose an extension of the reinforcement learning architecture that enables moral decision-making of reinforcement learning agents based on normative reasons. Central to this approach is a reason-based shield generator yielding a moral shield that binds the agent to actions that conform with recognized normative reasons so that our overall architecture restricts the agent to actions that are (internally) morally justified. In addition, we describe an algorithm that allows to iteratively improve the reason-based shield generator through case-based feedback from a moral judge.

cs.AI↗

arXiv2024

Enhancing GNNs with Architecture-Agnostic Graph Transformations: A Systematic Analysis

In recent years, a wide variety of graph neural network (GNN) architectures have emerged, each with its own strengths, weaknesses, and complexities. Various techniques, including rewiring, lifting, and node annotation with centrality values, have been employed as pre-processing steps to enhance GNN performance. However, there are no universally accepted best practices, and the impact of architecture and pre-processing on performance often remains opaque. This study systematically explores the impact of various graph transformations as pre-processing steps on the performance of common GNN architectures across standard datasets. The models are evaluated based on their ability to distinguish non-isomorphic graphs, referred to as expressivity. Our findings reveal that certain transformations, particularly those augmenting node features with centrality measures, consistently improve expressivity. However, these gains come with trade-offs, as methods like graph encoding, while enhancing expressivity, introduce numerical inaccuracies widely-used python packages. Additionally, we observe that these pre-processing techniques are limited when addressing complex tasks involving 3-WL and 4-WL indistinguishable graphs.

cs.LG↗

arXiv2024

Radiative transfer modeling of outbursts of massive young stellar objects

Young stellar objects (YSOs) accrete up to half of their material in short periods of enhanced mass accretion. For massive YSOs (MYSOs with more than 8 solar masses), accretion outbursts are of special importance, as they serve as diagnostics in highly obscured regions. Within this work, two outbursting MYSOs within different evolutionary stages, the young source G358.93-0.03 MM1 (G358) and the more evolved one G323.46-0.08 (G323), are investigated, and the major burst parameters are derived. For both sources, follow-up observations with the airborne SOFIA observatory were performed to detect the FIR afterglows. All together, we took three burst-/post-observations in the far infrared. The burst parameters are needed to understand the accretion physics and to conclude on the possible triggering mechanisms behind it. Up to today, G323s burst is the most energetic one ever observed for a MYSO. G358s burst was about two orders of magnitude weaker and shorter (2 months instead of 8 years). We suggest that G358s burst was caused by the accretion of a spiral fragment (or a small planet), where G323 accreted a heavy object (a planet or even a potential companion). To model those sources, we use radiative transfer (RT) simulations (static and time-dependent). G323s accretion burst is the first astrophysical science case, that is modeled with time-dependent RT (TDRT). We incorporate a small TDRT parameter-study and develop a time-depending fitting tool (the TFitter) for future modeling.

astro-ph.SR↗

arXiv2021

Variance Reduction in Stochastic Reaction Networks using Control Variates

Monte Carlo estimation in plays a crucial role in stochastic reaction networks. However, reducing the statistical uncertainty of the corresponding estimators requires sampling a large number of trajectories. We propose control variates based on the statistical moments of the process to reduce the estimators' variances. We develop an algorithm that selects an efficient subset of infinitely many control variates. To this end, the algorithm uses resampling and a redundancy-aware greedy selection. We demonstrate the efficiency of our approach in several case studies.

stat.ME↗

arXiv2021

GINA: Neural Relational Inference From Independent Snapshots

Dynamical systems in which local interactions among agents give rise to complex emerging phenomena are ubiquitous in nature and society. This work explores the problem of inferring the unknown interaction structure (represented as a graph) of such a system from measurements of its constituent agents or individual components (represented as nodes). We consider a setting where the underlying dynamical model is unknown and where different measurements (i.e., snapshots) may be independent (e.g., may stem from different experiments). We propose GINA (Graph Inference Network Architecture), a graph neural network (GNN) to simultaneously learn the latent interaction graph and, conditioned on the interaction graph, the prediction of a node's observable state based on adjacent vertices. GINA is based on the hypothesis that the ground truth interaction graph -- among all other potential graphs -- allows to predict the state of a node, given the states of its neighbors, with the highest accuracy. We test this hypothesis and demonstrate GINA's effectiveness on a wide range of interaction graphs and dynamical processes.

cs.LG↗

arXiv2021

Analysis of Markov Jump Processes under Terminal Constraints

Many probabilistic inference problems such as stochastic filtering or the computation of rare event probabilities require model analysis under initial and terminal constraints. We propose a solution to this bridging problem for the widely used class of population-structured Markov jump processes. The method is based on a state-space lumping scheme that aggregates states in a grid structure. The resulting approximate bridging distribution is used to iteratively refine relevant and truncate irrelevant parts of the state-space. This way the algorithm learns a well-justified finite-state projection yielding guaranteed lower bounds for the system behavior under endpoint constraints. We demonstrate the method's applicability to a wide range of problems such as Bayesian inference and the analysis of rare events.

eess.SY↗

arXiv2021

Abstraction-Guided Truncations for Stationary Distributions of Markov Population Models

To understand the long-run behavior of Markov population models, the computation of the stationary distribution is often a crucial part. We propose a truncation-based approximation that employs a state-space lumping scheme, aggregating states in a grid structure. The resulting approximate stationary distribution is used to iteratively refine relevant and truncate irrelevant parts of the state-space. This way, the algorithm learns a well-justified finite-state projection tailored to the stationary behavior. We demonstrate the method's applicability to a wide range of non-linear problems with complex stationary behaviors.

stat.ML↗

arXiv2021

Analytic solutions for stochastic hybrid models of gene regulatory networks

Discrete-state stochastic models are a popular approach to describe the inherent stochasticity of gene expression in single cells. The analysis of such models is hindered by the fact that the underlying discrete state space is extremely large. Therefore hybrid models, in which protein counts are replaced by average protein concentrations, have become a popular alternative. The evolution of the corresponding probability density functions is given by a coupled system of hyperbolic PDEs. This system has Markovian nature but its hyperbolic structure makes it difficult to apply standard functional analytical methods. We are able to prove convergence towards the stationary solution and determine such equilibrium explicitly by combining abstract methods from the theory of positive operators and elementary ideas from potential analysis.

math.AP↗

arXiv2020

Tracking the Race Between Deep Reinforcement Learning and Imitation Learning -- Extended Version

Learning-based approaches for solving large sequential decision making problems have become popular in recent years. The resulting agents perform differently and their characteristics depend on those of the underlying learning approach. Here, we consider a benchmark planning problem from the reinforcement learning domain, the Racetrack, to investigate the properties of agents derived from different deep (reinforcement) learning approaches. We compare the performance of deep supervised learning, in particular imitation learning, to reinforcement learning for the Racetrack model. We find that imitation learning yields agents that follow more risky paths. In contrast, the decisions of deep reinforcement learning are more foresighted, i.e., avoid states in which fatal decisions are more likely. Our evaluations show that for this sequential decision making problem, deep reinforcement learning performs best in many aspects even though for imitation learning optimal decisions are considered.

cs.LG↗

arXiv2020

Bounding Mean First Passage Times in Population Continuous-Time Markov Chains

We consider the problem of bounding mean first passage times for a class of continuous-time Markov chains that captures stochastic interactions between groups of identical agents. The quantitative analysis of such probabilistic population models is notoriously difficult since typically neither state-based numerical approaches nor methods based on stochastic sampling give efficient and accurate results. Here, we propose a technique that extends recently developed methods using semi-definite programming to determine bounds on mean first passage times. We further apply the technique to hybrid models and demonstrate its accuracy and efficiency for some examples from biology.

eess.SY↗

arXiv2019

Generalized Method of Moments Estimation for Stochastic Models of DNA Methylation Patterns

With recent advances in sequencing technologies, large amounts of epigenomic data have become available and computational methods are contributing significantly to the progress of epigenetic research. As an orthogonal approach to methods based on machine learning, mechanistic modeling aims at a description of the mechanisms underlying epigenetic changes. Here, we propose an efficient method for parameter estimation for stochastic models that describe the dynamics of DNA methylation patterns over time. Our method is based on the Generalized Method of Moments (GMM) and gives results with an accuracy similar to that of maximum likelihood-based estimation approaches. However, in contrast to the latter, the GMM still allows an efficient and accurate calibration of parameters even if the complexity of the model is increased by considering longer methylation patterns. We show the usefulness of our method by applying it to hairpin bisulfite sequencing data from mouse ESCs for varying pattern lengths.

q-bio.GN↗