arXiv subjects

Oliver Müller

Publications and source records attributed to Oliver Müller.

At least 19 recordsLinked to original sources

arXiv2026

A Dual-Process Perspective on Nudge Susceptibility in LLM-Based GUI Agents

LLM-based GUI agents increasingly act on behalf of users in digital environments that were designed with human users in mind. These graphical user interfaces were designed to support, but also deliberately steer, the behaviour and decisions of users. While behavioural biases in the textual outputs of LLMs are well-documented, far less is known about how such influence operates when models act as agents that perceive interfaces and execute decisions---and, in particular, whether the reasoning capabilities increasingly built into these agents make them more robust to it. Drawing on Dual-Process Theory, we empirically investigate whether LLM-based GUI agents are susceptible to automatic (Type 1) and reflective (Type 2) digital nudges, and how their reasoning configuration moderates this susceptibility. In a randomized online shopping experiment with 3,600 agents and a total of 21,600 simulations across six frontier models from three providers, we found that agents were vulnerable to both nudge types. Crucially, the reasoning configuration moderated these effects in opposing directions, reducing susceptibility to automatic default nudges while heightening it to reflective social influence nudges. Extensive reasoning therefore did not make agents more robust but redirected the route through which choice architecture takes effect. Exploratory analysis further showed this redirection to be systematically structured by model scale. Beyond establishing nudge susceptibility as a behavioural property of agentic AI, the study positions interface design as a governance concern for organizations that delegate decisions to autonomous agents.

cs.AI↗

arXiv2026

H-VAEP and H-xT: Valuing Offensive On-the-Ball Actions in Handball by Estimating Probabilities

Traditional player evaluation in professional handball relies on basic box-score metrics or heuristic indices, which fail to credit the multi-player build-up chain. While football (soccer) analytics has adopted Expected Threat (xT) and Valuing Actions by Estimating Probabilities (VAEP), these event-based action valuation frameworks have not yet been adapted to handball. In this paper, we present the first comprehensive adaptation and evaluation of xT and VAEP for handball, utilizing five seasons of tracking-derived event data from the Handball Bundesliga. We develop Handball-xT (H-xT) using a handball-native court zoning layout, demonstrating via simulations that it is systematically more robust than standard rectangular grids. We optimize Handball-VAEP (H-VAEP) by tailoring its feature space and selecting the context length to limit team-identity leakage. Our evaluation shows that H-VAEP yields exceptionally stable, discriminative, and intuitive player ratings that highlight build-up play. Finally, we release our complete code repository to help professional clubs deploy these models.

cs.LG↗

arXiv2026

Nudging Sustainable Choices through LLM-Generated Recommendation Explanations

Recommender systems mediate everyday consumption, offering a promising channel for encouraging sustainable choices. Prior research shows that explanations influence users' perceptions of recommendations and can support more informed decisions. We argue that explanations can also serve as behavioral nudges by foregrounding sustainability information at the moment of choice. This study investigates how different behavioral framings of sustainability information in recommendation explanations affect user choices and perceptions. Using generative AI, we generate sustainability-aware explanations by drawing on nudge theory and validate them through human evaluation and LLM-as-a-judge audits. Building on this foundation, we conduct two randomized studies (N = 529) in a low involvement domain (instant coffee) and a high involvement domain (hotel bookings), in which participants choose among preference matched recommendations accompanied by these explanations. Our results show that, across both domains, merely disclosing sustainability information in explanations does not change choices, whereas framing that information or invoking a descriptive social norm significantly increases sustainable selections and eases decision-making. Notably, perception and behavior diverge, as plain disclosure improves explanation evaluations without translating into more sustainable selection behavior. Our work demonstrates how LLMs can generate theory-grounded explanations at scale, pointing toward practical explanation-based interventions for social good. We conclude by discussing implications for adaptive explanation design with generative AI.

cs.AI↗

arXiv2026

LLM-SoccerArena: Benchmarking LLMs on Real-World Predictions in Sports

Large language models (LLMs) increasingly support decisions about uncertain future events, yet evaluating their ability to forecast real-world outcomes remains difficult. In particular, existing benchmarks are typically static and retrospective, and therefore cannot test how information is synthesized by LLMs to predict future events under uncertainty. We introduce LLM-SoccerArena (https://llm-soccerarena.com), a prospective live benchmark that evaluates how well LLMs forecast real-world sports events before the outcomes are known. LLM-SoccerArena provides (1) a prospective live benchmark protocol, (2) a public open-source platform, and (3) a factorial benchmark design together with tournament-related questions (e.g., which team will win). LLM-SoccerArena automatically records timestamped, schema-validated forecasts of unresolved events, together with prompts, model versions, tool traces, and costs. The factorial design varies along four dimensions: (1) model version (e.g., GPT-5.5, Claude Opus 4.8); (2) information access; (3) prompting strategy, and (4) forecast horizon. We demonstrate LLM-SoccerArena through a large-scale evaluation of the 2026 FIFA World Cup, in which seven LLMs generated forecasts for all 104 matches and 15 tournament-related questions. We provide a detailed analysis of model performance across information access, prompting strategy, and forecast horizon. As a result, LLM-SoccerArena provides new evidence about the forecasting performance of state-of-the-art LLMs. For example, LLMs with web access outperform those without, but only by a small margin (i.e., a 0.023 improvement in Brier score). Overall, LLM-SoccerArena provides a flexible, open-source platform for prospective benchmarking of unresolved events. LLM-SoccerArena will be continuously updated, and can be directly applied to future national and international tournaments and league competitions.

cs.AI↗

arXiv2026

Predicting Heterogeneous Treatment Effects Of Building Energy Saving Retrofits Using Causal Machine Learning

Information Systems research increasingly relies on machine learning (ML) to predict outcomes in complex sociotechnical systems, yet predictive models are not designed to identify causal effects. This limitation is particularly critical in building retrofits, where unbiased estimates of energy savings are essential for climate policy and investment decisions. Because retrofit adoption is shaped by household and building characteristics that also affect energy consumption, predictive ML can yield biased effect estimates. This paper systematically benchmarks leading causal ML estimators, including metalearners (S-, T- and X-Learners) and DoubleML across multiple retrofit interventions. To enable this comparison, we construct a physically grounded simulation in which true treatment effects and realistic adoption biases are known. Results show that DoubleML achieves the lowest estimation errors, particularly for complex envelope retrofits. These findings demonstrate that orthogonalising the treatment assignment improves causal effect estimation and provides a methodological foundation for large-scale energy retrofit and policy evaluation.

cs.CE↗

arXiv2026

TS-Arena -- A Live Forecast Pre-Registration Platform

Time Series Foundation Models (TSFMs) are transforming the field of forecasting. However, evaluating them on historical data is increasingly difficult due to the risks of train-test sample overlaps and temporal overlaps between correlated train and test time series. To address this, we introduce TS-Arena, a live forecasting platform that shifts evaluation from the known past to the unknown future. Building on the concept of continuous benchmarking, TS-Arena evaluates models on future data. Crucially, we introduce a strict forecasting pre-registration protocol: models must submit predictions before the ground-truth data physically exists. This makes test-set contamination impossible by design. The platform relies on a modular microservice architecture that harmonizes and structures data from different sources and orchestrates containerized model submissions. By enforcing a strict pre-registration protocol on live data streams, TS-Arena prevents information leakage offers a faster alternative to traditional static, infrequently repeated competitions (e.g. the M-Competitions). First empirical results derived from operating TS-Arena over one year of energy time series demonstrate that established TSFMs accumulate robust longitudinal scores over time, while the continuous nature of the benchmark simultaneously allows newcomers to demonstrate immediate competitiveness. TS-Arena provides the necessary infrastructure to assess the true generalization capabilities of modern forecasting models. The platform and corresponding code are available at https://ts-arena.live/.

cs.LG↗

arXiv2026

Rethinking Evaluation in the Era of Time Series Foundation Models: (Un)known Information Leakage Challenges

Time Series Foundation Models (TSFMs) represent a new paradigm for time-series forecasting, promising zero-shot predictions without the need for task-specific training or fine-tuning. However, similar to Large Language Models (LLMs), the evaluation of TSFMs is challenging: as training corpora grow increasingly large, it becomes difficult to ensure the integrity of the test sets used for benchmarking. An investigation of existing TSFM evaluation studies identifies two kinds of information leakage: (1) train-test sample overlaps arising from the multi-purpose reuse of datasets and (2) temporal overlap of correlated train and test series. Ignoring these forms of information leakage when benchmarking TSFMs risks producing overly optimistic performance estimates that fail to generalize to real-world settings. We therefore argue for the development of novel evaluation methodologies that avoid pitfalls already observed in both LLM and classical time-series benchmarking, and we call on the research community to adopt principled approaches to safeguard the integrity of TSFM evaluation.

cs.LG↗

arXiv2026

Detectability of Satellite Planes in Mock Observations of Isolated L* Galaxies

The existence and prevalence of planar, co-rotating distributions of satellite galaxies around L* host galaxies in the local universe remains a subject of ongoing debate. Despite numerous observational efforts over the past decade, a statistically robust sample of "satellite planes" across the diversity of host galaxy environments is lacking. To guide future observing strategies, we construct a controlled suite of mock observations of on-sky positions and line-of-sight (LOS) velocities of isolated L* host galaxies and their satellite systems, based on samples drawn from the Illustris TNG100-1 cosmological simulation to build a statistical sample. In these mock systems, satellite planes are defined by three key parameters: the number of satellites ($N_{\mathrm{sat}}$), the fraction residing in a thin co-rotating plane ($f_{p}$), and the orientation angle relative to the observer ($θ_{\mathrm{rot}}$). We evaluate the sensitivity of three observational metrics, $N_{\mathrm{cor}}$ (number of co-rotating satellites), $b/a$ (projected flattening of the satellite distribution), and $v_\mathrm{los}$ (mean absolute LOS velocity), to the presence of such planes. Our results show that detection rates are strongly dependent on $θ_{\mathrm{rot}}$ and $N_{\mathrm{sat}}$. Satellite planes that are viewed nearly edge-on or face-on, are the most readily detected. In contrast, intermediate orientations and systems with fewer satellites yield low detection success rates. Generally, only satellite planes with $N_{\mathrm{sat}}>20$ have high chances of being detected. These findings provide a practical framework for prioritising observational targets and designing future surveys aimed at detecting and characterising satellite planes.

astro-ph.GA↗

arXiv2026

Benchmarking Time Series Foundation Models for Short-Term Household Electricity Load Forecasting

Accurate household electricity short-term load forecasting (STLF) is key to future and sustainable energy systems. While various studies have analyzed statistical, machine learning, or deep learning approaches for household electricity STLF, recently proposed time series foundation models such as Chronos, TimesFM or Time-MoE promise a new approach for household electricity STLF. These models are trained on a vast amount of time series data and are able to forecast time series without explicit task-specific training (zero-shot learning). In this study, we benchmark the forecasting capabilities of time series foundation models compared to Trained-from-Scratch (TFS) Transformer-based approaches. Our results suggest that foundation models perform comparably to TFS Transformer models, while certain time series foundation models outperform all TFS models when the input size increases. At the same time, they require less effort, as they need no domain-specific training and only limited contextual data for inference.

cs.CE↗

arXiv2026

Distance measurements from the internal dynamics of globular clusters: Application to the Sombrero galaxy (M104)

Globular clusters (GCs) are dense star clusters found in all massive galaxies. Recent work has established that they follow a tight relation between their internal stellar velocity dispersion $σ$ and luminosity, enabling accurate distance measurements. In this work, we aim to apply this GC velocity dispersion (GCVD) distance method to measure the distance to M104 (NGC 4594, the Sombrero galaxy). We have measured internal stellar velocity dispersions for 85 globular clusters (GCs) and one ultra-compact dwarf galaxy around M104 using high-resolution multi-object integrated-light spectroscopy with FLAMES/GIRAFFE on the Very Large Telescope. The measured velocity dispersions range from $σ= 4 - 30$ km s$^{-1}$, with a mean uncertainty of $Δσ= 2.5$ km s$^{-1}$. For a subset of 77 GCs with $V$-band magnitudes and reliable velocity dispersion measurements above $σ> 4$ km s$^{-1}$, we constructed the $M_V$-$σ$ relation to measure the distance to M104, finding $D=9.00\pm0.29$ (stat.)~$\pm0.26$ (sys.) Mpc. The GCs follow the Milky Way and M31 $M_V-σ$ relation closely, with the exception of the luminous ultra-compact dwarf SUCD1, which is nearly one magnitude brighter than the mean relation. 29 GCs in the sample have sizes determined from Hubble Space Telescope imaging which allowed us to determine their masses and $V$-band dynamical mass-to-light ratios (M/L$_V$). We find a mean $ = 2.6 M_{\odot}/L_{\odot}$ for the luminous ($M_V < -8$ mag) M 104 GCs, which is higher than the Milky Way GCs, but is reminiscent of the brightest GCs in Centaurus A. With the exception of SUCD1, the GCs of M104 follow the GCVD relation irrespective of their mass-to-light ratio.

astro-ph.GA↗

arXiv2025

Dwarf galaxies in the MATLAS Survey: Hubble Space Telescope observations of nuclear star clusters

In dwarf galaxies, nuclear star clusters (NSCs) are believed to primarily form from the migration and merger of globular clusters (GCs), with a possible contribution from in-situ star-forming activity triggered by gas infall. We present the study of NSCs in 41 MATLAS survey dwarf galaxies including ultra-diffuse galaxies (UDGs), as part of a large follow-up imaging program with the Hubble Space Telescope (HST) Advanced Camera for Surveys (ACS) using the F606W and F814W filters. The sample is biased towards low-surface brightness and large dwarfs, i.e., UDG-like galaxies, and includes two galaxies with a double nucleus, 13 newly identified nucleated dwarfs thanks to HST's high spatial resolution, and five candidate ultra-compact dwarf progenitors. We modeled the NSCs with a Sérsic profile and derived their structural properties and photometry. We find the NSC Sérsic index to increase with the luminosity and stellar mass, while no obvious trend is seen on the effective radius and ellipticity. The faint NSCs tend to have a constant color profile, whereas the bright ones have a bluer center, suggesting that the most massive NSCs in our sample might have experienced a mixed formation scenario, including in-situ star formation. A significant portion of our NSCs tend to be more massive than for other galaxy samples of similar stellar mass, which could be due to some dwarfs ongoing tidal disruption or an initial formation of massive NSCs from multiple GC mergers and in-situ star forming activity. More observations of resolved NSC are needed to be able to infer their formation scenario from the structural properties and photometry in dwarfs.

astro-ph.GA↗

arXiv2025

Getting In Contract with Large Language Models -- An Agency Theory Perspective On Large Language Model Alignment

Adopting Large language models (LLMs) in organizations potentially revolutionizes our lives and work. However, they can generate off-topic, discriminating, or harmful content. This AI alignment problem often stems from misspecifications during the LLM adoption, unnoticed by the principal due to the LLM's black-box nature. While various research disciplines investigated AI alignment, they neither address the information asymmetries between organizational adopters and black-box LLM agents nor consider organizational AI adoption processes. Therefore, we propose LLM ATLAS (LLM Agency Theory-Led Alignment Strategy) a conceptual framework grounded in agency (contract) theory, to mitigate alignment problems during organizational LLM adoption. We conduct a conceptual literature analysis using the organizational LLM adoption phases and the agency theory as concepts. Our approach results in (1) providing an extended literature analysis process specific to AI alignment methods during organizational LLM adoption and (2) providing a first LLM alignment problem-solution space.

cs.AI↗

arXiv2025

Globular clusters in M104: Tracing kinematics and metallicities from the centre to the halo

As ancient star clusters, globular clusters (GCs) are regarded as powerful tracers of galaxy evolution and assembly. Due to their brightness and compact sizes, GCs are employed to probe the kinematics and stellar population properties of galaxies, from the central regions out into the halo where the underlying stellar light becomes too faint for spectroscopic studies. In this work, we present a comprehensive study of the GC system of M 104 (NGC 4594, also known as the Sombrero galaxy) based on literature spectroscopic catalogues and newly collected data from Very Large Telescope (VLT) MUSE integral-field spectroscopy combined with multi-object spectroscopy from VLT FLAMES and OSIRIS at the Gran Telescopio de Canarias (GTC). We present a new catalogue of 499 GCs with radial velocity measurements that span from the inner disc region out to $\sim$ 70 kpc (24$^{\prime}$). In addition to velocities, we measure metallicities from the MUSE, OSIRIS, and FLAMES spectra of 190 GCs. Together with literature values, we collected a sample of 278 metallicities. Comparing GCs observed with multiple instruments, we find a good agreement of velocity and metallicity measurements. Studying GC kinematics with a simple model confirms a decreasing velocity dispersion profile and low rotation velocities. The blue GCs appear to be more dispersion-dominated, while the red GCs follow the kinematics of the stars more closely. We find a large scatter of GC metallicities with distance from the centre and metal-rich GCs are found over all radii. We discuss that the GC metallicity distribution with a broad metal-poor component likely reflects the complex assembly history of M 104.

astro-ph.GA↗

arXiv2025

Low Surface Brightness structures from annotated deep CFHT images: effects of the host galaxy's properties and environment

Hierarchical galactic evolution models predict that mergers drive galaxy growth, producing low surface brightness (LSB) tidal features that trace galaxies' late assembly. These faint structures encode information about past mergers and are sensitive to the properties and environment of the host galaxy. We investigated the relationships between LSB features and their hosts in a sample of 475 nearby massive galaxies spanning diverse environments (field, groups, Virgo cluster) using deep optical imaging from the Canada-France-Hawaii Telescope (MATLAS, UNIONS/CFIS, VESTIGE, NGVS). Using Jafar, an online annotation tool, we manually annotated tidal features and extended stellar haloes, including 199 tidal tails and 100 streams. Geometric and photometric measurements were extracted to analyse their dependence on galaxy mass, environment, and internal kinematics. At our surface brightness limit of 29 mag$\,$arcsec$^{-2}$, tidal features and stellar haloes contribute 2% and 10% of total galaxy luminosity, respectively. Tidal features are detected in 36% of galaxies, with none fainter than 27.8 mag$\,$arcsec$^{-2}$. The most massive galaxies are twice as likely to host tidal debris, and for early-type galaxies their halos are twice as luminous as those in lower-mass systems, a trend not observed in late-type galaxies. Although small-scale interactions increase the frequency of tidal features, the large-scale environment does not influence it. An anticorrelation between this frequency and rotational support is found, but may reflect the mass-driven effect. We release our database of annotated features for deep learning applications. Our findings confirm that galaxy mass is the dominant factor influencing tidal feature prevalence, consistent with hierarchical formation models.

astro-ph.GA↗

arXiv2025

Galaxies OBserved as Low-luminosity Identified Nebulae (GOBLIN): a catalog of 43,000 high-probability dwarf galaxy candidates in the UNIONS survey

The detection of low surface brightness galaxies beyond the Local Group poses significant observational challenges, yet these faint systems are fundamental to our understanding of dark matter, hierarchical galaxy formation, and cosmic structure. Their abundance and distribution provide crucial tests for cosmological models, particularly regarding the small-scale predictions of $Λ$CDM. We present a systematic detection framework for dwarf galaxy candidates in Ultraviolet Near Infrared Optical Northern Survey (UNIONS) data covering 4,861 deg$^{2}$. Our pipeline preprocesses UNIONS gri-band data through binning, artifact removal, and stellar masking, then employs MTObjects (MTO) for low surface brightness detection. After parameter cuts and cross-matching, we obtain $\sim$360 candidates per deg$^{2}$, totaling $\sim$1.5 million candidates forming our GOBLIN (Galaxies OBserved as Low-luminosity Identified Nebulae) catalog. We fine-tuned the deep learning model Zoobot, pre-trained on Galaxy Zoo labels, for classification. Training data came from visual inspection of literature candidates with probability labels from expert assessments, capturing consensus and uncertainty. Applied to all MTO objects, our method identifies 42,965 dwarf candidates with probability $>$ 0.8, including 23,072 with probability $>$ 0.9. High-probability candidates correlate spatially with massive galaxies (log$(M_{*}/M_{\odot}) \geq$ 10) within 120 Mpc. While some of these objects may have been previously identified in other surveys, we present this extensive catalog of candidates, including their positions, structural parameter estimates, and classification probabilities, as a resource for the community to enable studies of galaxy formation, evolution, and the distribution of dwarf galaxies in different environments.

astro-ph.GA↗

arXiv2025

MUSE observations of dwarf galaxies and a stellar stream in the M83 group

Spectroscopy for faint dwarf galaxies outside of our own Local Group is challenging. Here, we present MUSE spectroscopy to study the properties of four known dwarf satellites and one stellar stream (KK208) surrounding the nearby grand spiral M83, which resides together with the lenticular galaxy Cen A in the Centaurus group. This data complete the phase-space information for all known dwarf galaxies around M83 down to a completeness of $-$10 mag in the $V$ band. All studied objects have an intermediate to old and metal-poor stellar population and follow the stellar luminosity-metallicity relation as defined by the Local Group dwarfs. For the stellar stream we serendipitously identify a previously unknown globular cluster, which is old and metal-poor. Two dwarf galaxies (NGC5264 and dw1341-29) may be a bound satellite of a satellite system due to their proximity and shared velocities. Having access to the positions and velocities of 13 dwarfs around M83, we estimate the mass of the group with different estimators. Ranging between 1.3 and $3.0 \times 10^{12}$ M$_\odot$ for the halo mass we find it to be larger than previously assumed. This may impact the previously reported tension for cold dark matter cosmology with the count of dwarf galaxies. In contrast to Cen A, we do not find a co-rotating plane-of-satellites around M83.

astro-ph.GA↗

arXiv2025

New dwarf galaxy candidates in the M106, NGC3521, and UGCA127 groups with the Hyper Suprime Camera

The local universe is still full of hidden dwarf galaxies to be discovered using deep imaging campaigns. Here we present the third paper in a series to search for low-surface brightness dwarf galaxies around nearby isolated luminous host galaxies with the Subaru Hyper Suprime Camera. Based on visual inspection, we found 11, 0, 4, and 6 dwarf galaxy candidates around M106, NGC2903, NGC3521, and UGCA127, respectively. This adds to the 40 candidates around M104 and 4 candidates around NGC2683 found in the previous papers. Artificial galaxy experiments show that we are complete down to a mean effective surface brightness of 26 mag/arcsec$^2$. The new dwarf galaxy candidates follow known scaling relation in size, surface brightness and luminosity, making them good candidates based on their morphology and photometric properties. We trace the luminosity function of these galaxies down to magnitude of $\approx-$9 in the V band for all galaxies targeted in our survey footprint so far. While the most massive galaxy (M104) has a significant higher abundance of dwarfs, NGC3521, NGC2903, and NGC2683 have a similar luminosity function as the Milky Way. These latter three galaxies also have a similar stellar mass and might be considered Milky Way analogs. UGCA127 is a low-mass galaxy but almost reaches the same number of dwarfs as the Milky Way at our limiting magnitude. We have searched for hints of lopsidedness in the satellite distributions, but found none to be significant. The next step will be to confirm these members through either distance or velocity measurements.

astro-ph.GA↗

arXiv2025

Evidence of star cluster migration and merger in dwarf galaxies

Nuclear star clusters (NSCs) are the densest stellar systems in the Universe. They can be found at the center of all galaxy types, but tend to favor galaxies of intermediate stellar mass around 10$^9\,$M$_{\odot}$[1, 2]. Currently, two main processes are under debate to explain their formation: in-situ star-formation from gas infall[3] and migration and merging of globular clusters (GCs) caused by dynamical friction[4]. Studies[5-9] of NSC stellar populations suggest that the former predominates in massive galaxies, the latter prevails in dwarf galaxies, and both contribute equally at intermediate mass. However, up to now, no ongoing merger of GCs has yet been observed to confirm this scenario. Here we report the serendipitous discovery of five dwarf galaxies with complex nuclear regions, characterized by multiple nuclei and tidal tails, using high resolution images from the Hubble Space Telescope. These structures have been reproduced in complementary N-body simulations, supporting the interpretation that they result from migrating and merging of star clusters. The small detection rate and short simulated timescales (below 100 Myr) of this process may explain why this has not been observed previously. This study highlights the need of large surveys with high resolution to fully map the migration scenario steps.

astro-ph.GA↗