arXiv subjects

Florian von Wangenheim

Publications and source records attributed to Florian von Wangenheim.

10 recordsLinked to original sources

arXiv2026

I Don't Miss You, but I Do: Self-Explanation Faithfulness of Modality Missingness in Vision-Language Models

Vision-language models (VLMs) are increasingly used in settings where some input modalities may be unavailable, yet we know little about whether they can faithfully explain how such missing information affects their own predictions. We introduce an interventional protocol for evaluating self-explanations of modality dynamics: models state what each modality alone would support, whether restoring a missing modality would change their answer, and whether the available evidence is sufficient; we then execute the corresponding intervention and compare these claims with realized behavior. We evaluate ten VLMs spanning open-weight and proprietary models across four tasks covering mixed, redundant, and unique modality regimes. We find a systematic tendency to overstate the sufficiency of available evidence. Models substantially underestimate the effect of restoring missing modalities: executed change exceeds predicted change in 78 of 80 model-task-condition settings, with task-level median executed change rates reaching 70.1\% while median predicted rates remain at most 9.6\%. Insufficiency claims have low recall, leaving many cases in which behavior changes despite a stated claim of sufficiency. Retrospective self-explanations show the same tendency, over-crediting single-input sufficiency in mixed regimes and interchangeability in redundant ones. Together, these results show that VLMs systematically mischaracterize how their predictions depend on available and missing evidence, motivating executable interventions as a behavioral test of multimodal self-explanations.

cs.LG↗

arXiv2026

When, Not How Much: Evaluating Time-Series Foundation Models on Sparse Events

Pretrained time-series foundation models (TSFMs) are evaluated as forecasters of future values, yet for sparse series many decisions depend only on which future periods contain activity. Standard benchmarks do not assess this. On five sparse datasets, we rank positions within forecast windows that contain both events and zeros. The released point forecasts of 12 TSFMs improve chance-corrected average precision over training-free references by at most 0.031, and in chance-corrected AUC the median TSFM falls below them on every dataset. With event supervision, linear probes of six frozen backbones improve on their backbone's point forecast in 29 of 30 backbone--dataset pairs. Averaging the predicted quantiles instead of taking their median improves the ranking of most TSFMs that forecast the median, and on two datasets the strongest such outputs rival the probes. The probes' advantage over raw-context learners depends on the dataset, and under the same probe, pretrained features outperform randomly initialized ones for five of six backbones. For sparse-event ranking, released point forecasts thus add little over simple references, whereas lightweight event heads on frozen TSFMs rank events better than these forecasts, and the best of them exceed gradient-boosted trees trained on the raw context on three of the five datasets. More broadly, assessing pretrained forecasters on tasks beyond value forecasting requires reporting their outputs, supervised probes of their representations, and raw-context and randomized controls side by side, since each supports a different conclusion.

cs.LG↗

arXiv2026

NeuReasoner: Theory-grounded Mapping of Reasoning Elicitation Boundaries

A growing body of work suggests that the reasoning capabilities of large language models are largely latent in their base form, with post-training primarily amplifying rather than introducing them. However, this evidence comes mainly from mathematical and coding benchmarks, leaving the boundary conditions of that claim largely unexplored, namely which cognitive tasks can be recovered through elicitation and where that recovery fails. To investigate this, we introduce NeuReasoner, a theory-grounded elicitation instrument. At each step, an orchestrator pairs a Neuro Lens, inspired by functional specificity, with a Cognitive Lens, drawn from the Erotetic Theory of Reasoning, and integrates their outputs through internal modularization of a single model, without external tools. We evaluate NeuReasoner on CogBench, a suite of behavioral tasks from cognitive psychology, alongside standard mathematical and coding benchmarks, measuring both its improvement over vanilla inference and its ability to match a model's post-trained thinking mode. At sufficient scale, NeuReasoner matches or exceeds thinking-mode baselines on arithmetic reasoning, code generation, Bayesian reasoning, and reward learning; these gains persist against self-consistency and iterative-refinement baselines matched to NeuReasoner's per-decision call budget. Using NeuReasoner allows us to find clear boundaries: risk-taking and decision making under uncertainty remains hard to recover through elicitation alone, and model scale interacts with elicitation in both directions: widening its advantage on some cognitive signatures while erasing it on others. Overall, through NeuReasoner as a modular, interpretable, theory-grounded elicitation instrument, we empirically map where reasoning elicitation succeeds and fails, beyond the mathematical and coding benchmarks where prior claims have rested.

cs.LG↗

arXiv2026

MESSY STREETS: A Benchmark for Geocoding Real-World Addresses

We introduce MESSY STREETS, a benchmark for evaluating geocoders on verbatim web addresses, with existence verification and controlled measurement of surface-form divergence. Unlike conventional benchmarks based on clean or synthetically perturbed addresses, MESSY STREETS contains addresses whose surface forms diverge from canonical representations and whose components may be missing, repeated, malformed, or incomplete. The benchmark is constructed from the December 2024 Web Data Commons corpus, with reference locations established from OpenAddresses or OpenStreetMap. The strongest commercial geocoders outperform open-source systems by up to 49 percentage points in recall. This gap is driven primarily by differences in candidate return rates on non-canonical addresses; once a candidate is returned, positional accuracy is broadly comparable across systems. Non-canonical surface form alone accounts for up to 25 percentage points of recall loss. Examining Nominatim's query-processing pipeline, we show that its conjunctive matching lets a single unrecognised token zero an otherwise valid query. The results demonstrate that geocoder choice is a consequential design decision for applications processing noisy address data, and that normalisation and preprocessing could substantially narrow the gap between open-source and commercial geocoders.

cs.IR↗

arXiv2026

Detecting Drunk Driving Using Off-the-Shelf Smartwatches

Alcohol-impaired driving remains a major yet preventable cause of road traffic injury and death, with many drivers underestimating their level of intoxication. Compared to in-vehicle systems, mobile drunk-driving detection using consumer smartwatches offers a scalable way to trigger preventive interventions and increase awareness without additional in-vehicle hardware. We introduce a system that leverages wrist accelerometer data and heart rate variability-derived physiological signals to detect alcohol-related driving impairment. We collected data in a randomized, controlled three-arm test-track study (n=54) and trained both logistic regression models with window-aggregated features and a two-tower 1D convolutional neural network (CNN), to detect alcohol-impaired driving. The CNN achieved a participant-averaged area under the receiver operating characteristic (AUROC) of 0.88 for detecting any alcohol intoxication and 0.86 for detecting driving above the WHO-recommended limit of 0.05 g/dL. To the best of our knowledge, this is the first work to (1) demonstrate drunk-driving detection using consumer smartwatches, (2) develop and evaluate such a system in a real vehicle on a closed test track, and (3) rigorously assess generalization to unseen participants. Together, these findings highlight the potential of wearable-based sensing to support scalable, measurement-driven prevention of alcohol-related traffic harm.

cs.HC↗

arXiv2026

EventChat: Implementation and user-centric evaluation of a large language model-driven conversational recommender system for exploring leisure events in an SME context

Large language models (LLMs) present an enormous evolution in the strategic potential of conversational recommender systems (CRS). Yet to date, research has predominantly focused upon technical frameworks to implement LLM-driven CRS, rather than end-user evaluations or strategic implications for firms, particularly from the perspective of a small to medium enterprises (SME) that makeup the bedrock of the global economy. In the current paper, we detail the design of an LLM-driven CRS in an SME setting, and its subsequent performance in the field using both objective system metrics and subjective user evaluations. While doing so, we additionally outline a short-form revised ResQue model for evaluating LLM-driven CRS, enabling replicability in a rapidly evolving field. Our results reveal good system performance from a user experience perspective (85.5% recommendation accuracy) but underscore latency, cost, and quality issues challenging business viability. Notably, with a median cost of $0.04 per interaction and a latency of 5.7s, cost-effectiveness and response time emerge as crucial areas for achieving a more user-friendly and economically viable LLM-driven CRS for SME settings. One major driver of these costs is the use of an advanced LLM as a ranker within the retrieval-augmented generation (RAG) technique. Our results additionally indicate that relying solely on approaches such as Prompt-based learning with ChatGPT as the underlying LLM makes it challenging to achieve satisfying quality in a production environment. Strategic considerations for SMEs deploying an LLM-driven CRS are outlined, particularly considering trade-offs in the current technical landscape.

cs.IR↗

arXiv2026

Pre-Clinical Latency Characterization of VRxBioRelax: A Real-Time EMG Biofeedback System for Muscle Relaxation in Virtual Reality

Chronic tension in the upper trapezius (UT), often caused by poor ergonomics, prolonged posture, or psychological stress, contributes to musculoskeletal discomfort, headaches, and impaired interoceptive awareness. Although surface electromyography (sEMG) biofeedback can promote UT relaxation, traditional systems using conventional displays often fail to sustain engagement. Virtual reality (VR) offers a more immersive alternative, provided that latency remains below perceptual thresholds. We introduce VRxBioRelax, a closed-loop VR biofeedback system that streams sEMG data from Delsys Trigno Avanti sensors via MQTT to a Unity scene. Muscle activation drives a dynamic dawn-to-dusk landscape synchronized with a progressive muscle relaxation protocol. To validate system responsiveness, 87,716 EMG samples from the NinaPro DB2 dataset were replayed at $\sim$75 Hz. Timestamps at four key stages-acquisition, Root Mean Square (RMS) processing, network receipt, and rendering-revealed mean latencies of 0.50 ms (processing), 5.62 ms (network), and 19.22 ms (rendering), yielding an average end-to-end delay of 25.34 ms. Notably, 99.3% of frames arrived within 50 ms. One-sided t-tests confirmed mean latency was significantly lower than both the 30 ms VR comfort limit ($t_{87\,715}=-25.2$, $p=5.9{\times}10^{-140}$) and the 50 ms clinical benchmark ($t_{87\,715}=-133.3$, $p<10^{-300}$). These findings support VRxBioRelax for use in remote interoceptive training, stress reduction, and telepresence-enabled rehabilitation.

cs.HC↗

arXiv2026

Detecting and Mitigating Group Bias in Heterogeneous Treatment Effects

Heterogeneous treatment effects (HTEs) are increasingly estimated using machine learning models that produce highly personalized predictions of treatment effects. In practice, however, predicted treatment effects are rarely interpreted, reported, or audited at the individual level but, instead, are often aggregated to broader subgroups, such as demographic segments, risk strata, or markets. We show that such aggregation can induce systematic bias of the group-level causal effect: even when models for predicting the individual-level conditional average treatment effect (CATE) are correctly specified and trained on data from randomized experiments, aggregating the predicted CATEs up to the group level does not, in general, recover the corresponding group average treatment effect (GATE). We develop a unified statistical framework to detect and mitigate this form of group bias in randomized experiments. We first define group bias as the discrepancy between the model-implied and experimentally identified GATEs, derive an asymptotically normal estimator, and then provide a simple-to-implement statistical test. For mitigation, we propose a shrinkage-based bias-correction, and show that the theoretically optimal and empirically feasible solutions have closed-form expressions. The framework is fully general, imposes minimal assumptions, and only requires computing sample moments. We analyze the economic implications of mitigating detected group bias for profit-maximizing personalized targeting, thereby characterizing when bias correction alters targeting decisions and profits, and the trade-offs involved. Applications to large-scale experimental data at major digital platforms validate our theoretical results and demonstrate empirical performance.

stat.ME↗

arXiv2025

RAxSS: Retrieval-Augmented Sparse Sampling for Explainable Variable-Length Medical Time Series Classification

Medical time series analysis is challenging due to data sparsity, noise, and highly variable recording lengths. Prior work has shown that stochastic sparse sampling effectively handles variable-length signals, while retrieval-augmented approaches improve explainability and robustness to noise and weak temporal correlations. In this study, we generalize the stochastic sparse sampling framework for retrieval-informed classification. Specifically, we weight window predictions by within-channel similarity and aggregate them in probability space, yielding convex series-level scores and an explicit evidence trail for explainability. Our method achieves competitive iEEG classification performance and provides practitioners with greater transparency and explainability. We evaluate our method in iEEG recordings collected in four medical centers, demonstrating its potential for reliable and explainable clinical variable-length time series classification.

cs.LG↗

arXiv2022

The Value of Measuring Trust in AI - A Socio-Technical System Perspective

Building trust in AI-based systems is deemed critical for their adoption and appropriate use. Recent research has thus attempted to evaluate how various attributes of these systems affect user trust. However, limitations regarding the definition and measurement of trust in AI have hampered progress in the field, leading to results that are inconsistent or difficult to compare. In this work, we provide an overview of the main limitations in defining and measuring trust in AI. We focus on the attempt of giving trust in AI a numerical value and its utility in informing the design of real-world human-AI interactions. Taking a socio-technical system perspective on AI, we explore two distinct approaches to tackle these challenges. We provide actionable recommendations on how these approaches can be implemented in practice and inform the design of human-AI interactions. We thereby aim to provide a starting point for researchers and designers to re-evaluate the current focus on trust in AI, improving the alignment between what empirical research paradigms may offer and the expectations of real-world human-AI interactions.

cs.HC↗