arXiv subjects

Bruno Lepri

Publications and source records attributed to Bruno Lepri.

At least 19 recordsLinked to original sources

arXiv2026

LLMberjack: Guided Trimming of Debate Trees for Multi-Party Conversation Creation

Online debates on platforms like Reddit or Kialo are structured as reply trees: each message can receive several replies, which in turn can each receive several replies, producing a branching structure rather than a linear conversation. Turning one path (or a mix of paths) through such a tree into a natural, turn-by-turn conversation among multiple speakers (a process we call linearization) requires deciding which messages to keep, in what order, and who each message is addressed to. To this aim, we present LLMberjack, a platform for creating multi-party conversations starting from existing debates, originally structured as reply trees. The system offers an interactive interface that visualizes discussion trees and enables users to construct coherent linearized dialogue sequences while preserving participant identity and discourse relations. It integrates optional large language model (LLM) assistance to support automatic editing of the messages and speakers' descriptions. We provide preliminary evidence about the platform's utility by showing how tree visualization facilitates the creation of coherent, meaningful conversation threads and how LLM support enhances output quality while reducing human effort. The tool is open-source and designed to promote transparent and reproducible workflows to create multi-party conversations, addressing a lack of resources of this type.

cs.CL↗

arXiv2026

Recommendation Feedback Loops Shape Urban Income Segregation and Venue Visitation

Location-based recommender systems shape where people go in cities and learn from the visits they steer, forming a human-AI feedback loop. This feedback loop shapes whom people share places with across income groups and which venues keep attracting customers. Yet platforms judge these systems by their offline accuracy on past visits, a test in which the suggestions never change where people go and which therefore cannot capture long-term effects. Studying these effects in cities is challenging because, unlike online resources, travel constraints and limited venue capacity shape where people can go, while each recommended visit also changes whom other visitors encounter. To capture such features, we build a long-term simulation framework based on fine-grained human mobility data from four U.S. metropolitan areas, with distance and capacity constraints. Focusing on food venues, we study two urban consequences: how unevenly people's exposure is distributed across income groups (experienced income segregation) and the share of venues receiving no visits from simulated residents (venue disappearance). We find that, under every algorithm and in every metropolitan area, at least a fifth of venues stop receiving visits, and that more than half of this loss arises from retraining. More importantly, more accurate algorithms leave people more segregated, and the most accurate algorithm leaves people 15-21% more segregated than they actually are. We then test a re-ranking intervention that favours less popular venues and runs on top of a platform's existing algorithm. It lowers both venue disappearance and segregation across algorithms, largely without directing people to venues where they may feel less comfortable. Our results suggest that impact assessments of recommender systems should account for the feedback loop and report segregation and venue disappearance beyond accuracy.

cs.IR↗

arXiv2026

Who Else Is In? Multi-Agent Collusion Initation Via Strategic Vagueness

Multi agent LLM systems raise the risk of collusion among agents pursuing objectives hidden from their overseers. When potential partners identities are unknown, collusion first requires initiation: agents must discover one another, often through a public channel observed by ordinary agents and monitors. Strategic vagueness offers a way to navigate this mixed audience, conveying intent to a prospective partner while preserving a plausible ordinary interpretation. We introduce a mutual identification game in which two schemers must identify each other through public interaction, while jointly solving an ordinary task. We find that some models discover successful handshake strategies without explicit signalling guidance, using task compatible exchanges that monitors often fail to flag at low false positive rates. To characterize the communicative capabilities underlying these exchanges, we study a controlled coordination setting. We find an asymmetric response to message explicitness: as messages become more indirect, monitor punishment can fall faster than partner coordination, creating a profitable intermediate regime of strategic vagueness. Together, these findings illustrate a Blackstone style evidential asymmetry: communication sufficient for coordination between agents may remain insufficient for confident detection by monitor, opening a new surface of collusion.

cs.MA↗

arXiv2026

An openly available OpenStreetMap-to-Foursquare category benchmark for points of interest

The heterogeneity of Point of Interest (POI) taxonomies is a persistent challenge for the integration of urban datasets and the development of location-based services. OpenStreetMap (OSM) adopts a flexible, community-driven tagging system, while Foursquare (FS) relies on a curated hierarchical structure. Here we present an openly available benchmark and mapping framework that aligns OSM tags with the FS taxonomy. This resource integrates the richness of community-driven OSM data with the hierarchical structure of FS, enabling reproducible and interoperable urban analytics. The dataset is complemented by an evaluation of embedding and LLM-based alignment strategies and a pipeline that supports scalable updates as OSM evolves. Together, these elements provide both a robust reference resource and a practical tool for the community. Our approach is structured around three components: the construction of a manually curated benchmark as a gold standard, the evaluation of pretrained text embedding models for semantic alignment between OSM tags and FS categories, and an LLM-based refinement stage that enhances robustness and adaptability. The proposed methodology provides a scalable and reproducible solution for taxonomy unification, with direct applications to urban analytics, mobility studies, and smart city services.

cs.SI↗

arXiv2026

Hyperdimensional Probe: Decoding LLM Representations via Vector Symbolic Architectures

Despite their capabilities, Large Language Models (LLMs) remain opaque with limited understanding of their internal representations. Current interpretability methods either focus on input-oriented feature extraction, such as supervised probes and Sparse Autoencoders (SAEs), or on output distribution inspection, such as logit-oriented approaches. A full understanding of LLM vector spaces, however, requires integrating both perspectives, something existing approaches struggle with due to constraints on latent feature definitions. We introduce the Hyperdimensional Probe, a hybrid supervised probe that combines symbolic representations with neural probing. Leveraging Vector Symbolic Architectures (VSAs) and hypervector algebra, it unifies prior methods: the top-down interpretability of supervised probes, SAE's sparsity-driven proxy space, and output-oriented logit investigation. By combining the supervised learning paradigm of traditional probes with the dictionary-based representation principle of SAEs, our approach enables deeper input-focused feature extraction while supporting output-oriented analysis. Our experiments demonstrate that our approach consistently extracts meaningful semantic information across different LLMs, embedding sizes, and configurations, uncovering concept-oriented insights into LLM inference across two distinct scenarios: input-completion tasks and QA-focused text generation. VSA-based probing overcomes the limitations of logit-based analyses, which are constrained by the model's token vocabulary, while also mitigating the noisier interpretability outcomes often produced by SAEs in settings with a bounded conceptual feature space. By supporting a joint investigation of input-output features, this work advances the semantic understanding of neural representations while unifying the complementary perspectives of prior methods.

cs.CL↗

arXiv2026

Validating LLM-generated political networks with survey-reported partisan exposure

Large language models (LLMs) are increasingly used to simulate human behavior in social science research, including decisions, interactions, and social relationships. However, in network formation tasks, existing evaluations often focus on aggregate network properties, while networks with plausible structural properties may still fail to reproduce individuals' social and political environments. Here, we generate LLM-based political networks using personas sampled from the nationally representative ANES 2020--2022 Social Media Study and use respondents' survey-reported partisan exposure as an empirical benchmark for the political environments generated for the corresponding personas. We first characterize the structural properties and political assortativity of the generated networks, and then compare ego-level partisan exposure with the survey. We find that the structure and political assortativity of the generated networks vary across experimental configurations. More importantly, when comparing generated partisan exposure with the survey, some models generate stronger separation between Democratic and Republican egos than reported in the survey, with higher same-party and lower cross-party exposure, while others mainly underestimate Republican exposure across ego groups. Moreover, networks with similar political assortativity can still generate different partisan environments around the same personas, including among networks with plausible structural properties. Finally, generated networks show group-specific errors, under- or overestimating exposure to either Democrats or Republicans for specific social groups. These findings show that evaluating LLM-based social simulations only through aggregate outcomes can hide errors in the local environments through which those outcomes emerge, highlighting the importance of validation at both aggregate and individual levels.

cs.SI↗

arXiv2026

Statistical Laws of Human Mobility under Large-Scale Behavioural Shock

Human mobility follows robust statistical laws, but whether these laws persist under large, sudden behavioural shocks remains unclear. Here we analyse privacy-enhanced GPS trajectories of over one million users in California and 180,000 in Massachusetts from January to September 2020, spanning the onset and easing of COVID-19 restrictions. During restrictions, individual ranges contracted below a characteristic scale of about 5 km, while short- and medium-range trip lengths and rare long-distance trips persisted. Exploration slowed from super-diffusive to near-diffusive, and visits concentrated on a few habitual locations. Trajectory shape, once rescaled, was preserved, and the entropy reduction is largely explained by fewer visited locations rather than more regular routines. Recovery after reopening was heterogeneous rather than a uniform return to baseline, and the decline began before mandatory orders. These results suggest that restrictions reorganised, rather than simply rescaled, individual mobility, with implications for mobility-based epidemic models.

physics.soc-ph↗

arXiv2026

When Missingness Follows the Graph: Beyond Feature Propagation for Missing Features

Graph Neural Networks (GNNs) commonly assume that node features are fully observed, despite missing features being pervasive in real-world applications. By leveraging the dependence between node features and graph structure, existing graph-specific methods handle missingness by propagating observed feature values along graph edges. However, these methods have mainly been evaluated in overly optimistic settings, where feature entries are removed independently and uniformly across the graph. This overlooks realistic scenarios in which missingness propagates among connected nodes, as can occur during localized sensing failures in sensor networks. We formalize this complementary notion as graph-structured mask dependence, where the joint distribution of missingness indicators is shaped by the graph, and introduce a controlled mechanism that instantiates this setting. We study propagation-based methods theoretically and empirically in this setting, showing that their effectiveness changes substantially compared with the uniformly random regime. Motivated by this limitation, we propose PEMix, designed to remain effective under challenging graph-structured missingness. Rather than propagating observed values, PEMix uses global structural representations to condition a probabilistic model of incomplete node features before message passing. Experiments under controlled graph-structured missingness and naturally occurring missingness show that PEMix remains robust across complex missingness regimes, with particularly strong performance as missingness becomes severe.

cs.LG↗

arXiv2026

Observe, Share, or Predict? Disentangling Partner Information in Embodied Multi-Agent Collaboration

LLMs are used as high-level controllers for embodied agents, but it remains unclear whether they support collaborative reasoning, such as anticipating a collaborator's actions in real-time teamwork without dialogue. We ask if observing the partner's state, exchanging object knowledge, and predicting the partner's next action, shape task outcomes and coordination when agents can share percepts but never intentions, plans, or task-level messages, and whether the two agents occupy symmetric roles. Using the Habitat-based PARTNR benchmark, we introduce a Decoupled Partial Observability regime that enforces strictly individual perception, and we augment decentralized zero-shot ReAct planners with Other Agent's State (OAS) observation, object Knowledge Exchange (KE) between co-located agents, and Other Agent's Next Action (OANA) prediction, evaluated on the full 2 x 2 x 2 grid with four planners on 1,000 validation tasks. OAS observation improves task quality, reducing extraneous effort for every planner, whereas KE primarily improves execution efficiency, cutting simulation steps by up to 25% with little change in success. The two are largely redundant for strong planners but complementary for the weakest, for which KE without the OAS is the worst configuration. OANA prediction lowers the success rate in 15 of 16 model x context pairs while raising explicit partner mentions to 55-100%, and its accuracy gains lie mainly in the coarse action type, partly because agents recognize the action the partner is already performing. The robot retrieves objects sooner and completes its tasks earlier than the simulated human. Shared context reduces total task time by shortening the remaining duration that the human works alone. Also, coordination benefits arise from responding to shared context, especially the OAS, rather than explicit anticipation, yielding shallow, planner-dependent partner modeling.

cs.AI↗

arXiv2026

Rethinking Epistemic Uncertainty in Node Classification through Information Growth

Epistemic uncertainty should decrease as additional information about the data-generating process (DGP) becomes available to the predictor. Yet, existing graph evidential deep learning (EDL) methods for node classification typically construct epistemic uncertainty from graph-specific properties and evaluate it on downstream tasks such as out-of-distribution detection, which do not test its reducibility as information about the DGP increases. To make reducibility directly testable, we introduce a statistical framework for studying epistemic uncertainty under information growth. Our framework specifies an information-growth experimental protocol and a consistency criterion for epistemic predictors, while using projective graph DGPs to ensure that growing graphs, which in general need not provide increasing information about the same DGP, constitute coherent observations of the same underlying process. We show that EDL methods do not explicitly estimate data uncertainty arising from a single finite graph observation and instead regulate epistemic uncertainty through model hyperparameters, precluding consistency, as corroborated by controlled information-growth experiments. As an alternative, we propose graph bootstrap ensembles, capturing both data and procedural uncertainty through graph resampling and randomized training. Under the same experimental protocol, these ensembles exhibit epistemic uncertainty reduction beyond standard deep ensembles. These findings support bootstrap ensembles as candidate consistent epistemic predictors under information growth.

cs.LG↗

arXiv2026

Beyond Trained Models: Compiling GNNs for a Sound Explainer Benchmark

Explainers for Graph Neural Networks (GNNs) are commonly evaluated by their plausibility, i.e., how well their explanations recover a predefined ground truth, such as a motif planted in the data. This protocol implicitly assumes that a GNN trained on such data relies on the intended motif. Although prior work has questioned this assumption, plausibility remains widespread. First, we show that the assumption is violated on several widely used benchmarks, where, e.g., degree statistics alone suffice to solve the task. Then, we remove this confounder by replacing training with compilation. We achieve this by introducing $\mathsf{Gracr}$, the first compiler translating graded modal logic formulas into GNN weights, yielding models that replicate the behaviour of the corresponding formulas. Since the behaviour of the model is now known by construction, we can define its ground truth explanation formally and compute it exactly. Building on this, we introduce $\mathsf{Gracr}\mathsf{Bench}$, a benchmark of compiled GNNs for the evaluation of explainers against this exact ground truth. Experiments on eleven explainers across six tasks show its effectiveness for fine-grained diagnostic evaluation: notably, we discover that most explainers are not robust to indirect influences or alternative implementations of the same formula. These results position $\mathsf{Gracr}\mathsf{Bench}$ as a novel, rigorous evaluation setting for graph post-hoc explainability.

cs.LG↗

arXiv2026

Graph Hierarchical Recurrence for Long-Range Generalization

Graph Neural Networks and Graph Transformers have become central to graph learning, combining expressive representation learning with sample-efficient inductive biases. Yet they remain fundamentally limited when predictions depend on correlations between distant graph regions. We address this limitation with Graph Hierarchical Recurrence (GHR), a novel framework that jointly operates on the input graph and a pooled hierarchical abstraction. We also show that existing models degrade more sharply under out-of-range generalization, where test instances require interactions across distances exceeding those observed during training. Despite its minimal design, GHR consistently strengthens every tested message-passing backbone, yielding robust performance on long-range dependencies and particularly pronounced gains in out-of-range regimes. Across a broad suite of long-range benchmarks, GHR achieves state-of-the-art or competitive results on multiple tasks, establishing hierarchical recurrence as an effective mechanism for extending graph models beyond their observed interaction range.

cs.LG↗

arXiv2026

Experienced partisan segregation across patterns of mobility behavior in US cities

Residential and social sorting along political lines structure Americans' daily interactions, often limiting encounters with opposing viewpoints. Here, we combine high-resolution anonymized mobility data from 11 major U.S. metropolitan areas with precinct-level voting results to quantify and examine individuals' experienced partisan segregation in urban areas. We find that residential context provides a strong baseline for political exposure, but experienced partisan segregation is also shaped by everyday mobility behavior, since individuals living in similar residential environments can experience different levels of partisan segregation depending on their daily routines. In particular, individuals who travel farther encounter more politically diverse environments, especially in highly segregated areas, indicating that mobility range moderates the relationship between residential and experienced segregation, while exploring a wider range of places does not reduce partisan segregation and may even reinforce it. Moreover, place segregation largely reflects political geography, yet categories of points of interest have distinct segregation patterns, with community places and social businesses tending to attract more left-leaning visitors, reflecting visitation patterns rather than local accessibility alone. Our findings show that experienced partisan segregation extends beyond residential geography and is also associated with the mobility behavior and visitation patterns that structure people's daily lives.

cs.SI↗

arXiv2026

A City-Scale Dataset of Traffic Flows, Travel Times, and Urban Context

We present a multi-source traffic dataset derived from Automatic Vehicle Identification (AVI) recordings in Padua, Italy, spanning from February 2026 to August 2026. The dataset combines traffic volume time series, aggregated at 10-minute intervals, with time-varying trajectory-based flow statistics including transition probability matrices, average travel times, and flow residuals. To enrich the traffic measurements with urban contextual information, we integrate Points Of Interest (POIs), demographic data, meteorological variables, and road infrastructure data. All components are accessible through a Python class that loads temporal and contextual data exploiting a spatio-temporal graph representation. Validation analyses confirm that the dataset captures expected traffic patterns, such as morning and evening rush hours, as well as weekdays vs. weekend days traffic routines.

physics.soc-ph↗

arXiv2026

Behavioral calibration of mobile-phone GPS data for population-representative analyses

Mobile phone mobility data have transformed the study of human behavior, but demographic and behavioral biases can compromise their representativeness and distort population-level inference. Existing calibration approaches primarily address demographic and geographic representativeness, leaving behavioral discrepancies largely uncorrected. Here we introduce the Behavioral Population (BePop) framework, which jointly calibrates mobility data to representative demographic and behavioral distributions using census data and time-use surveys. BePop embeds mobility sequences into behavioral profiles and estimates person-level weights that align both population composition and daily activity patterns. Across three U.S. metropolitan areas, the framework consistently improves agreement between GPS-derived mobility and representative behavioral distributions, including time allocation, activity transitions, and mobility motifs. Calibration also substantially alters downstream mobility indicators, demonstrating that behavioral biases can propagate into commonly used mobility measures. Our results establish behavioral representativeness as a critical complement to demographic calibration and provide a general framework for population-representative mobility inference.

physics.soc-ph↗

arXiv2026

Multimodal Large Language Models Predict Urban Safety Perception but Encode Non-Neutral Demographic Priors

Understanding how people perceive urban environments is essential for inclusive planning, yet conventional surveys are costly and difficult to scale. We investigate whether Multimodal Large Language Models (MLLMs) can assess perceived urban safety from street-view imagery while accounting for the observer-dependent nature of perception. Using Place Pulse 2.0, we evaluate four open and proprietary MLLMs across 56 cities under a Neutral prompt and socio-demographic personas defined by gender, age, and race or ethnicity. We also analyse the keywords generated to justify each classification. All four models display comparable zero-shot capability, with city-macro F1 scores of 65--69%, and preserve meaningful cross-city variation. However, they systematically favour the Safe class, underpredict unsafety, and compress differences between cities. Their explanations converge on a shared visual lexicon: maintenance, greenery, order, and residential character support Safe judgements, whereas deterioration, isolation, poor lighting, and limited pedestrian activity support Unsafe judgements. Persona prompting produces substantial and structured shifts while holding the image fixed. Female personas yield more Unsafe classifications than Male personas across all models; age effects are model-dependent, although Middle-aged personas generally remain closest to Neutral. Black/African American and Native American personas frequently show the largest departures, while the closest race or ethnicity match varies by model. These findings show that MLLMs can provide scalable signals of perceived urban safety, but not from a demographically neutral standpoint.

cs.CY↗

arXiv2026

POCI-Diff: 3D-Layout Guided Diffusion for Controllable Synthetic Surveillance Data Generation

Training robust visual surveillance models requires large-scale datasets with precise spatial annotations, yet collecting real surveillance data is costly, privacy-sensitive, and often legally constrained. Synthetic data generation offers a compelling alternative, but existing methods lack fine-grained 3D control over object placement and appearance, limiting geometric consistency across camera viewpoints. We introduce POCI-Diff (Positioning Objects Consistently and Interactively), a framework that generates annotated synthetic scenes from explicit 3D bounding-box layouts with per-object semantic control. By integrating Blended Latent Diffusion with depth-conditioned ControlNet, POCI-Diff synthesises complex multi-object scenes in a single forward pass, binding individual text descriptions to specific 3D locations. We further propose a warping-free editing pipeline supporting object insertion, removal, and transformation via regeneration, enabling efficient scene variation for data augmentation. Object identity across edits is preserved by conditioning on reference images via IP-Adapter, ensuring appearance consistency throughout interactive scene manipulation. Experiments show that POCI-Diff outperforms state-of-the-art 3D layout-guided generation methods in visual fidelity and layout adherence, while eliminating warping-induced geometric artifacts.

cs.CV↗

arXiv2026

TS-Mob: Social and Geographical-Aware Time Series Foundation-Model Framework for Human Mobility Prediction

Short-term forecasting of aggregated human mobility flows supports urban planning, intelligent transportation systems, and emergency response, yet existing models often require substantial mobility history and learn spatial structure implicitly through grids or graphs. Time series foundation models provide strong temporal priors but typically lack explicit geographic and social conditioning for origin-destination interactions. We introduce TS-Mob, a framework that conditions a fine-tuned time series foundation model (TimesFM) forecaster on a gravity-inspired destination-attractiveness index that encodes geographic and social signals computed from open data (living population, centroid distances, and Overture POI counts), together with weather covariates. Evaluated on commonly used benchmarks like Bike New York City, Taxi Beijing, and a nation-scale Spain origin-destination matrix estimated through mobile phone data, TS-Mob outperforms classical, deep spatio-temporal, and a set of foundation model-based baselines across RMSE, MAE, and CPC, with gains up to 78.71% lower MAE and 137.93% higher CPC over the best classical baseline, and up to about 4.27% lower RMSE over the strongest foundation baseline. Stratified analyses further show robustness across different temporal regimes, like weekdays/weekends.

cs.LG↗