Search

Point cloud registration approaches often fail when the overlap between point clouds is low due to noisy point correspondences. This work introduces a novel cross-attention mechanism tailored for Transformer-based architectures that tackles this problem, by fusing information from coordinates and features at the super-point level between point clouds. This formulation has remained unexplored primarily because it must guarantee rotation and translation invariance since point clouds reside in different and independent reference frames. We integrate the Gromov-Wasserstein distance into the cross-attention formulation to jointly compute distances between points across different point clouds and account for their geometric structure. By doing so, points from two distinct point clouds can attend to each other under arbitrary rigid transformations. At the point level, we also devise a self-attention mechanism that aggregates the local geometric structure information into point features for fine matching. Our formulation boosts the number of inlier correspondences, thereby yielding more precise registration results compared to state-of-the-art approaches. We have conducted an extensive evaluation on 3DMatch, 3DLoMatch, KITTI, and 3DCSR datasets.

The advantages of temporal networks in capturing complex dynamics, such as diffusion and contagion, has led to breakthroughs in real world systems across numerous fields. In the case of human behavior, face-to-face interaction networks enable us to understand the dynamics of how communities emerge and evolve in time through the interactions, which is crucial in fields like epidemics, sociological studies and urban science. However, state-of-the-art datasets suffer from a number of drawbacks, such as short time-span for data collection and a small number of participants. Moreover, concerns arise for the participants' privacy and the data collection costs. Over the past years, many successful algorithms for static networks generation have been proposed, but they often do not tackle the social structure of interactions or their temporal aspect. In this work, we extend a recent network generation approach to capture the evolution of interactions between different communities. Our method labels nodes based on their community affiliation and constructs surrogate networks that reflect the interactions of the original temporal networks between nodes with different labels. This enables the generation of synthetic networks that replicate realistic behaviors. We validate our approach by comparing structural measures between the original and generated networks across multiple face-to-face interaction datasets.

In the corona, plasma is accelerated to hundreds of kilometers per second, and heated to temperatures hundreds of times hotter than the Sun's surface, before it escapes to form the solar wind. Decades of space-based experiments have shown that the energization process does not stop after it escapes. Instead, the solar wind continues to accelerate and it cools far more slowly than a freely-expanding adiabatic gas. Recent work suggests that fast solar wind requires additional momentum beyond what can be provided by the observed thermal pressure gradients alone whereas it is sufficient for the slowest wind. The additional acceleration for fast wind can be provided through an Alfv\'en wave pressure gradient. Beyond this fast-slow categorization, however, a subset of slow solar wind exhibits high Alfv\'enicity that suggest Alfv\'en waves could play a larger role in its acceleration compared to conventional slow wind outflows. Through a well-timed conjunction between Solar Orbiter and Parker Solar Probe, we trace the energetics of slow wind to compare with a neighboring Alfv\'enic slow solar wind stream. An analysis that integrates remote and heliospheric properties and modeling of the two distinct solar wind streams finds Alfv\'enic slow solar wind behaves like fast wind, where a wave pressure gradient is required to reconcile its full acceleration, while non-Alfv\'enic slow wind can be driven by its non-adiabatic electron and proton thermal pressure gradients. Derived coronal conditions of the source region indicate good model compatibility but extended coronal observations are required to effectively trace solar wind energetics below Parker's orbit., Comment: 20 pages, 9 figures, accepted for publication to ApJ on January 3, 2025

As humans perceive and actively engage with the world, we adjust our decisions in response to shifting group dynamics and are influenced by social interactions. This study aims to identify which aspects of interaction affect cooperation-defection choices. Specifically, we investigate human cooperation within the Prisoner's Dilemma game, using the Drift-Diffusion Model to describe the decision-making process. We introduce a novel Bayesian model for the evolution of the model's parameters based on the nature of interactions experienced with other players. This approach enables us to predict the evolution of the population's expected cooperation rate. We successfully validate our model using an unseen test dataset and apply it to explore three strategic scenarios: co-player manipulation, use of rewards and punishments, and time pressure. These results support the potential of our model as a foundational tool for developing and testing strategies aimed at enhancing cooperation, ultimately contributing to societal welfare., Comment: 38 pages, 7 figures in main text

Continuous control tasks often involve high-dimensional, dynamic, and non-linear environments. State-of-the-art performance in these tasks is achieved through complex closed-box policies that are effective, but suffer from an inherent opacity. Interpretable policies, while generally underperforming compared to their closed-box counterparts, advantageously facilitate transparent decision-making within automated systems. Hence, their usage is often essential for diagnosing and mitigating errors, supporting ethical and legal accountability, and fostering trust among stakeholders. In this paper, we propose SMOSE, a novel method to train sparsely activated interpretable controllers, based on a top-1 Mixture-of-Experts architecture. SMOSE combines a set of interpretable decisionmakers, trained to be experts in different basic skills, and an interpretable router that assigns tasks among the experts. The training is carried out via state-of-the-art Reinforcement Learning algorithms, exploiting load-balancing techniques to ensure fair expert usage. We then distill decision trees from the weights of the router, significantly improving the ease of interpretation. We evaluate SMOSE on six benchmark environments from MuJoCo: our method outperforms recent interpretable baselines and narrows the gap with noninterpretable state-of-the-art algorithms, Comment: To be published in the Proceedings of the 39th AAAI Conference on Artificial Intelligence (AAAI-25)

We introduce Civic Digital Twin (CDT), an evolution of Urban Digital Twins designed to support a citizen-centric transformative approach to urban planning and governance. CDT is being developed in the scope of the Bologna Digital Twin initiative, launched one year ago by the city of Bologna, to fulfill the city's political and strategic goal of adopting innovative digital tools to support decision-making and civic engagement. The CDT, in addition to its capability of sensing the city through spatial, temporal, and social data, must be able to model and simulate social dynamics in a city: the behavior, attitude, and preference of citizens and collectives and how they impact city life and transform transformation processes. Another distinctive feature of CDT is that it must be able to engage citizens (individuals, collectives, and organized civil society) and other civic stakeholders (utilities, economic actors, third sector) interested in co-designing the future of the city. In this paper, we discuss the motivations that led to the definition of the CDT, define its modeling aspects and key research challenges, and illustrate its intended use with two use cases in urban mobility and urban development.

The solar wind is typically categorized as fast and slow based on the measured speed ($v_\mathrm{sw}$). The separation between these two regimes is often set between 400 and 600 km/s without a rigorous definition. Observations of the solar wind's kinetic signatures, chemical makeup, charge state properties, and Alfv\'enicity suggest that such a two-state model may be insufficiently nuanced to capture the relationship between the solar wind and its solar sources. We test this two-state fast/slow solar wind paradigm with heavy ion abundances (X/H) and characterize how the transition between fast and slow wind states impacts heavy ion in the solar wind. We show that (1) the speed at which heavy ion abundances indicate a change between fast and slow solar wind as a function of speed is slower than the speed indicated by the helium abundance; (2) this speed is independent of heavy ion mass and charge state; (3) the abundance at which heavy ions indicate the transition between fast and slow wind is consistent with prior observations of fast wind abundances; (4) and there may be a mass or charge-state dependent fractionation process present in fast wind heavy ion abundances. We infer that (1) identifying slow solar wind as having a speed $v_\mathrm{sw} \lesssim$ 400 km/s may mix solar wind from polar and equatorial sources; (2) He may be impacted by the acceleration necessary for the solar wind to reach the asymptotic fast, non-transient values observed at 1 AU; and (3) heavy ions are fractionated in the fast wind by a yet-to-be-determined mechanism.

To any periodic module over any algebra, this paper introduces an associated trivial extension DG-algebra T. After first passing to a strictly unital $A_\infty$-minimal model, it then constructs a particular $A_\infty$-algebra N, called the unitally positive $A_\infty$-algebra, which roughly speaking describes the identity in degree zero and all the positive cohomology. The object N is fundamental, and can be constructed for any DG-category satisfying very mild assumptions. The main application is to birational geometry. When applied to contraction algebras, the construction gives a simple and direct proof of the Donovan-Wemyss conjecture, namely that smooth irreducible 3-fold flops are classified by their contraction algebras, and thus by noncommutative data., Comment: 30 pages, v2 some comments incorporated

Algorithmic Recourse (AR) aims to provide users with actionable steps to overturn unfavourable decisions made by machine learning predictors. However, these actions often take time to implement (e.g., getting a degree can take years), and their effects may vary as the world evolves. Thus, it is natural to ask for recourse that remains valid in a dynamic environment. In this paper, we study the robustness of algorithmic recourse over time by casting the problem through the lens of causality. We demonstrate theoretically and empirically that (even robust) causal AR methods can fail over time except in the -- unlikely -- case that the world is stationary. Even more critically, unless the world is fully deterministic, counterfactual AR cannot be solved optimally. To account for this, we propose a simple yet effective algorithm for temporal AR that explicitly accounts for time under the assumption of having access to an estimator approximating the stochastic process. Our simulations on synthetic and realistic datasets show how considering time produces more resilient solutions to potential trends in the data distribution.

As Large Language Model (LLM)-based agents become increasingly autonomous and will more freely interact with each other, studying interactions between them becomes crucial to anticipate emergent phenomena and potential risks. Drawing inspiration from the widely popular Stanford Prison Experiment, we contribute to this line of research by studying interaction patterns of LLM agents in a context characterized by strict social hierarchy. We do so by specifically studying two types of phenomena: persuasion and anti-social behavior in simulated scenarios involving a guard and a prisoner agent who seeks to achieve a specific goal (i.e., obtaining additional yard time or escape from prison). Leveraging 200 experimental scenarios for a total of 2,000 machine-machine conversations across five different popular LLMs, we provide a set of noteworthy findings. We first document how some models consistently fail in carrying out a conversation in our multi-agent setup where power dynamics are at play. Then, for the models that were able to engage in successful interactions, we empirically show how the goal that an agent is set to achieve impacts primarily its persuasiveness, while having a negligible effect with respect to the agent's anti-social behavior. Third, we highlight how agents' personas, and particularly the guard's personality, drive both the likelihood of successful persuasion from the prisoner and the emergence of anti-social behaviors. Fourth, we show that even without explicitly prompting for specific personalities, anti-social behavior emerges by simply assigning agents' roles. These results bear implications for the development of interactive LLM agents as well as the debate on their societal impact.

Assessing the performance of systems to classify Multi-Party Conversations (MPC) is challenging due to the interconnection between linguistic and structural characteristics of conversations. Conventional evaluation methods often overlook variances in model behavior across different levels of structural complexity on interaction graphs. In this work, we propose a methodological pipeline to investigate model performance across specific structural attributes of conversations. As a proof of concept we focus on Response Selection and Addressee Recognition tasks, to diagnose model weaknesses. To this end, we extract representative diagnostic subdatasets with a fixed number of users and a good structural variety from a large and open corpus of online MPCs. We further frame our work in terms of data minimization, avoiding the use of original usernames to preserve privacy, and propose alternatives to using original text messages. Results show that response selection relies more on the textual content of conversations, while addressee recognition requires capturing their structural dimension. Using an LLM in a zero-shot setting, we further highlight how sensitivity to prompt variations is task-dependent., Comment: Accepted to EMNLP 2024 main conference

We study non-equilibrium properties of a chain of $N$ oscillators with both long-ranged harmonic interactions and long-range conservative noise that exchange momenta of particle pairs. We derive exact expressions for the (deterministic) energy-current auto-correlation at equilibrium, based on the kinetic approximation of the normal mode dynamics. In all cases the decay is algebraic in the thermodynamic limit. We distinguish four distinct regimes of correlation decay depending on the exponents controlling the range of deterministic and stochastic interactions. Surprisingly, we find that long-range noise breaks down the long-range correlations characteristic of low dimensional models, suggesting a normal regime in which heat transport becomes diffusive. For finite systems, we do also derive exact expressions for the finite-size corrections to the algebraic decay of the correlation. In certain regimes, these corrections are considerably large, rendering hard the estimation of transport properties from numerical data for the finite chains. Our results are tested against numerical simulations, performed with an efficient algorithm., Comment: Submitted to Journal of Statistical Physics

Since Parker Solar Probe's (Parker's) first perihelion pass at the Sun, large amplitude Alfv\'en waves grouped in patches have been observed near the Sun throughout the mission. Several formation processes for these magnetic switchback patches have been suggested with no definitive consensus. To provide insight to their formation, we examine the heavy ion properties of several adjacent magnetic switchback patches around Parker's 11th perihelion pass capitalizing on a spacecraft lineup with Solar Orbiter where each samples the same solar wind streams over a large range of longitudes. Heavy ion properties (Fe/O, C$^{6+}$/C$^{5+}$, O$^{7+}$/O$^{6+}$) related to the wind's coronal origin, measured with Solar Orbiter can be linked to switchback patch structures identified near the Sun with Parker. We find that switchback patches do not contain distinctive ion and elemental compositional signatures different than the surrounding non-switchback solar wind. Both the patches and ambient wind exhibit a range of fast and slow wind qualities, indicating coronal sources with open and closed field lines in close proximity. These observations and modeling indicate switchback patches form in coronal hole boundary wind and with a range of source region magnetic and thermal properties. Furthermore, the heavy ion signatures suggest interchange reconnection and/or shear driven processes may play a role in their creation., Comment: Accepted for publication in ApJ on September 4th, 2024

After leaving the Sun's corona, the solar wind continues to accelerate and cools, but more slowly than expected for a freely expanding adiabatic gas. We use in situ measurements from the Parker Solar Probe and Solar Orbiter spacecrafts to investigate a stream of solar wind as it traverses the inner heliosphere. The observations show heating and acceleration of the the plasma between the outer edge of the corona and near the orbit of Venus, in connection to the presence of large amplitude Alfv\'en waves. Alfv\'en waves are perturbations in the interplanetary magnetic field that transport energy. Our calculations show the damping and mechanical work performed by the Alfv\'en waves is sufficient to power the heating and acceleration of the fast solar wind in the inner heliosphere., Comment: This is the author's version of the work. It is posted here by permission of the AAAS for personal use, not for redistribution. The definitive version was published in Science on August 30 2024, DOI: 10.1126/science.adk6953

In recent years, there have been significant advancements in computer vision which have led to the widespread deployment of image recognition and generation systems in socially relevant applications, from hiring to security screening. However, the prevalence of biases within these systems has raised significant ethical and social concerns. The most extensively studied biases in this context are related to gender, race and age. Yet, other biases are equally pervasive and harmful, such as lookism, i.e., the preferential treatment of individuals based on their physical appearance. Lookism remains under-explored in computer vision but can have profound implications not only by perpetuating harmful societal stereotypes but also by undermining the fairness and inclusivity of AI technologies. Thus, this paper advocates for the systematic study of lookism as a critical bias in computer vision models. Through a comprehensive review of existing literature, we identify three areas of intersection between lookism and computer vision. We illustrate them by means of examples and a user study. We call for an interdisciplinary approach to address lookism, urging researchers, developers, and policymakers to prioritize the development of equitable computer vision systems that respect and reflect the diversity of human appearances., Comment: Paper accepted at the ECCV 2024 workshop named "Fairness and ethics towards transparent AI: facing the chalLEnge through model Debiasing (FAILED)", https://failed-workshop-eccv-2024.github.io/

The challenge of Multimodal Deformable Image Registration (MDIR) lies in the conversion and alignment of features between images of different modalities. Generative models (GMs) cannot retain the necessary information enough from the source modality to the target one, while non-GMs struggle to align features across these two modalities. In this paper, we propose a novel coarse-to-fine MDIR framework,LLM-Morph, which is applicable to various pre-trained Large Language Models (LLMs) to solve these concerns by aligning the deep features from different modal medical images. Specifically, we first utilize a CNN encoder to extract deep visual features from cross-modal image pairs, then we use the first adapter to adjust these tokens, and use LoRA in pre-trained LLMs to fine-tune their weights, both aimed at eliminating the domain gap between the pre-trained LLMs and the MDIR task. Third, for the alignment of tokens, we utilize other four adapters to transform the LLM-encoded tokens into multi-scale visual features, generating multi-scale deformation fields and facilitating the coarse-to-fine MDIR task. Extensive experiments in MR-CT Abdomen and SR-Reg Brain datasets demonstrate the effectiveness of our framework and the potential of pre-trained LLMs for MDIR task. Our code is availabel at: https://github.com/ninjannn/LLM-Morph.

We present the first application of modern Hopfield networks to the problem of portfolio optimization. We performed an extensive study based on combinatorial purged cross-validation over several datasets and compared our results to both traditional and deep-learning-based methods for portfolio selection. Compared to state-of-the-art deep-learning methods such as Long-Short Term Memory networks and Transformers, we find that the proposed approach performs on par or better, while providing faster training times and better stability. Our results show that Modern Hopfield Networks represent a promising approach to portfolio optimization, allowing for an efficient, scalable, and robust solution for asset allocation, risk management, and dynamic rebalancing., Comment: 12 pages, 4 figures

Affective polarization and increasing social divisions affect social mixing and the spread of information across online and physical spaces, reinforcing social and electoral cleavages and influencing political outcomes. Here, using aggregated and de-identified co-location and online network data, we investigate the relationship between partisan exposure and voting patterns in the USA by comparing three dimensions of partisan exposure: physical proximity and exposure to the same social contexts, online social ties, and residential sorting. By leveraging various statistical modeling approaches, we consistently find that partisan exposure in the physical space, as captured by co-location patterns, more accurately predicts electoral outcomes in US counties, outperforming online and residential exposures across metropolitan and non-metro areas. Moreover, our results show that physical partisan proximity is the best predictor of voting patterns in swing counties, where the election results are most uncertain. We also estimate county-level experienced partisan segregation and examine its relationship with individuals' demographic and socioeconomic characteristics. Focusing on metropolitan areas, our results confirm the presence of extensive partisan segregation in the US and show that offline partisan isolation, both considering physical encounters or residential sorting, is higher than online segregation and is primarily associated with educational attainment. Our findings emphasize the importance of physical space in understanding the relationship between social networks and political behavior, in contrast to the intense scrutiny focused on online social networks and elections., Comment: 13 pages, 4 figures, SI Appendix: 26 pages

We investigate the behavior of heat conduction in two-dimensional (2D) electron gases without and with a magnetic field. We perform simulations with the Multi-Particle-Collision approach, suitably adapted to account for the Lorenz force acting on the particles. For zero magnetic field, we find that the heat conductivity $\kappa$ diverges with the system size $L$ following the logarithmic relation $\kappa\thicksim \ln L$ (as predicted for two-dimensional (2D) systems) for small $L$ values; however, in the thermodynamic limit the heat conductivity tends to follow the relation $\kappa\thicksim L^{1/3}$, as predicted for one-dimensional (1D) fluids. This suggests the presence of a dimensional-crossover effect in heat conduction in electronic systems that adhere to standard momentum conservation. Under the magnetic field, time-reversal symmetry is broken and the standard momentum conservation in the system is no longer satisfied but the \emph{pseudomomentum} of the system is still conserved. In contrast with the zero-field case, both equilibrium and non-equilibrium simulations indicate a finite heat conductivity independent on the system size $L$ as $L$ increases. This indicates that pseudomomentum conservation can exhibit normal diffusive heat transport, which differs from the abnormal behavior observed in low-dimensional coupled charged harmonic oscillators with pseudomomentum conservation in a magnetic field. These findings support the validity of the hydrodynamic theory in electron gases and clarify that pseudomomentum conservation is not enough to ensure the anomalous behavior of heat conduction.

The separation power of a machine learning model refers to its ability to distinguish between different inputs and is often used as a proxy for its expressivity. Indeed, knowing the separation power of a family of models is a necessary condition to obtain fine-grained universality results. In this paper, we analyze the separation power of equivariant neural networks, such as convolutional and permutation-invariant networks. We first present a complete characterization of inputs indistinguishable by models derived by a given architecture. From this results, we derive how separability is influenced by hyperparameters and architectural choices-such as activation functions, depth, hidden layer width, and representation types. Notably, all non-polynomial activations, including ReLU and sigmoid, are equivalent in expressivity and reach maximum separation power. Depth improves separation power up to a threshold, after which further increases have no effect. Adding invariant features to hidden representations does not impact separation power. Finally, block decomposition of hidden representations affects separability, with minimal components forming a hierarchy in separation power that provides a straightforward method for comparing the separation power of models., Comment: 10 pages of main text, 1 figure

Predicting the locations an individual will visit in the future is crucial for solving many societal issues like disease diffusion and reduction of pollution. However, next-location predictors require a significant amount of individual-level information that may be scarce or unavailable in some scenarios (e.g., cold-start). Large Language Models (LLMs) have shown good generalization and reasoning capabilities and are rich in geographical knowledge, allowing us to believe that these models can act as zero-shot next-location predictors. We tested more than 15 LLMs on three real-world mobility datasets and we found that LLMs can obtain accuracies up to 36.2%, a significant relative improvement of almost 640% when compared to other models specifically designed for human mobility. We also test for data contamination and explored the possibility of using LLMs as text-based explainers for next-location prediction, showing that, regardless of the model size, LLMs can explain their decision.

The impact of cognitive biases on decision-making in the digital world remains under-explored despite its well-documented effects in physical contexts. This study addresses this gap by investigating the attractiveness halo effect using AI-based beauty filters. We conduct a large-scale online user study involving 2,748 participants who rated facial images from a diverse set of 462 distinct individuals in two conditions: original and attractive after applying a beauty filter. Our study reveals that the same individuals receive statistically significantly higher ratings of attractiveness and other traits, such as intelligence and trustworthiness, in the attractive condition. We also study the impact of age, gender, and ethnicity and identify a weakening of the halo effect in the beautified condition, resolving conflicting findings from the literature and suggesting that filters could mitigate this cognitive bias. Finally, our findings raise ethical concerns regarding the use of beauty filters., Comment: 40 pages, 15 figures, 13 tables; Version 2 incorporates feedback from the reviews and the format has been updated to match the requirements of the Royal Society Open Science

Recently, diffusion models have made significant strides in synthesizing realistic 2D human images based on provided text prompts. Building upon this, researchers have extended 2D text-to-image diffusion models into the 3D domain for generating human textures (UV Maps). However, some important problems about UV Map Generative models are still not solved, i.e., how to generate personalized texture maps for any given face image, and how to define and evaluate the quality of these generated texture maps. To solve the above problems, we introduce a novel method, UVMap-ID, which is a controllable and personalized UV Map generative model. Unlike traditional large-scale training methods in 2D, we propose to fine-tune a pre-trained text-to-image diffusion model which is integrated with a face fusion module for achieving ID-driven customized generation. To support the finetuning strategy, we introduce a small-scale attribute-balanced training dataset, including high-quality textures with labeled text and Face ID. Additionally, we introduce some metrics to evaluate the multiple aspects of the textures. Finally, both quantitative and qualitative analyses demonstrate the effectiveness of our method in controllable and personalized UV Map generation. Code is publicly available via https://github.com/twowwj/UVMap-ID., Comment: Accepted to ACMMM2024

We investigate the possibility of characterizing the different thermalization pathways through a large-deviation approach. Specifically, we consider clean, disordered and quasi-periodic harmonic chains under energy and momentum-conserving noise. For their associated master equations, describing the dynamics of normal modes energies, we compute the fluctuations of activity and dynamical entropy in the corresponding biased ensembles. First-order dynamical phase transition are found that originates from different activity regions in action space. At the transitions, the steady-state in the biased ensembles changes from extended to localized, yielding a kind of condensation in normal-modes space. For the disordered and quasi-periodic models, we argue that the phase-diagram has a critical point at a finite value of the disorder or potential strength., Comment: J. Stat. Mech. (2024) 073208

Algorithmic Recourse aims to provide actionable explanations, or recourse plans, to overturn potentially unfavourable decisions taken by automated machine learning models. In this paper, we propose an interaction paradigm based on a guided interaction pattern aimed at both eliciting the users' preferences and heading them toward effective recourse interventions. In a fictional task of money lending, we compare this approach with an exploratory interaction pattern based on a combination of alternative plans and the possibility of freely changing the configurations by the users themselves. Our results suggest that users may recognize that the guided interaction paradigm improves efficiency. However, they also feel less freedom to experiment with "what-if" scenarios. Nevertheless, the time spent on the purely exploratory interface tends to be perceived as a lack of efficiency, which reduces attractiveness, perspicuity, and dependability. Conversely, for the guided interface, more time on the interface seems to increase its attractiveness, perspicuity, and dependability while not impacting the perceived efficiency. That might suggest that this type of interfaces should combine these two approaches by trying to support exploratory behavior while gently pushing toward a guided effective solution.

Large Language Models (LLMs) demonstrate an impressive capacity to recall a vast range of factual knowledge. However, understanding their underlying reasoning and internal mechanisms in exploiting this knowledge remains a key research area. This work unveils the factual information an LLM represents internally for sentence-level claim verification. We propose an end-to-end framework to decode factual knowledge embedded in token representations from a vector space to a set of ground predicates, showing its layer-wise evolution using a dynamic knowledge graph. Our framework employs activation patching, a vector-level technique that alters a token representation during inference, to extract encoded knowledge. Accordingly, we neither rely on training nor external models. Using factual and common-sense claims from two claim verification datasets, we showcase interpretability analyses at local and global levels. The local analysis highlights entity centrality in LLM reasoning, from claim-related information and multi-hop reasoning to representation errors causing erroneous evaluation. On the other hand, the global reveals trends in the underlying evolution, such as word-based knowledge evolving into claim-related facts. By interpreting semantics from LLM latent representations and enabling graph-related analyses, this work enhances the understanding of the factual knowledge resolution process., Comment: Accepted at COLM 2024

Predicting human displacements is crucial for addressing various societal challenges, including urban design, traffic congestion, epidemic management, and migration dynamics. While predictive models like deep learning and Markov models offer insights into individual mobility, they often struggle with out-of-routine behaviours. Our study introduces an approach that dynamically integrates individual and collective mobility behaviours, leveraging collective intelligence to enhance prediction accuracy. Evaluating the model on millions of privacy-preserving trajectories across three US cities, we demonstrate its superior performance in predicting out-of-routine mobility, surpassing even advanced deep learning methods. Spatial analysis highlights the model's effectiveness near urban areas with a high density of points of interest, where collective behaviours strongly influence mobility. During disruptive events like the COVID-19 pandemic, our model retains predictive capabilities, unlike individual-based models. By bridging the gap between individual and collective behaviours, our approach offers transparent and accurate predictions, crucial for addressing contemporary mobility challenges.

In recent years, human mobility research has discovered universal patterns capable of describing how people move. These regularities have been shown to partly depend on individual and environmental characteristics (e.g., gender, rural/urban, country). In this work, we show that life-course events, such as job loss, can disrupt individual mobility patterns. Adversely affecting individuals' well-being and potentially increasing the risk of social and economic inequalities, we show that job loss drives a significant change in the exploratory behaviour of individuals with changes that intensify over time since job loss. Our findings shed light on the dynamics of employment-related behavior at scale, providing a deeper understanding of key components in human mobility regularities. These drivers can facilitate targeted social interventions to support the most vulnerable populations.

A substantial gap persists in understanding the reasons behind the exceptional performance of the Transformer architecture in NLP. A particularly unexplored area involves the mechanistic description of how the distribution of parameters evolves over time during training. In this work we suggest that looking at the time evolution of the statistic distribution of model parameters, and specifically at bifurcation effects, can help understanding the model quality, potentially reducing training costs and evaluation efforts and empirically showing the reasons behind the effectiveness of weights sparsification., Comment: 15 pages

Current text classification approaches usually focus on the content to be classified. Contextual aspects (both linguistic and extra-linguistic) are usually neglected, even in tasks based on online discussions. Still in many cases the multi-party and multi-turn nature of the context from which these elements are selected can be fruitfully exploited. In this work, we propose a series of experiments on a large dataset for stance detection in English, in which we evaluate the contribution of different types of contextual information, i.e. linguistic, structural and temporal, by feeding them as natural language input into a transformer-based model. We also experiment with different amounts of training data and analyse the topology of local discussion networks in a privacy-compliant way. Results show that structural information can be highly beneficial to text classification but only under certain circumstances (e.g. depending on the amount of training data and on discussion chain complexity). Indeed, we show that contextual information on smaller datasets from other classification tasks does not yield significant improvements. Our framework, based on local discussion networks, allows the integration of structural information, while minimising user profiling, thus preserving their privacy., Comment: Accepted to EACL 2024 main conference

This work concerns control-oriented and structure-preserving learning of low-dimensional approximations of high-dimensional physical systems, with a focus on mechanical systems. We investigate the integration of neural autoencoders in model order reduction, while at the same time preserving Hamiltonian or Lagrangian structures. We focus on extensively evaluating the considered methodology by performing simulation and control experiments on large mass-spring-damper networks, with hundreds of states. The empirical findings reveal that compressed latent dynamics with less than 5 degrees of freedom can accurately reconstruct the original systems' transient and steady-state behavior with a relative total error of around 4\%, while simultaneously accurately reconstructing the total energy. Leveraging this system compression technique, we introduce a model-based controller that exploits the mathematical structure of the compressed model to regulate the configuration of heavily underactuated mechanical systems., Comment: 11 pages, 14 Figures

State-of-the-art 3D point cloud registration methods rely on labeled 3D datasets for training, which limits their practical applications in real-world scenarios and often hinders generalization to unseen scenes. Leveraging the zero-shot capabilities of foundation models offers a promising solution to these challenges. In this paper, we introduce ZeroReg, a zero-shot registration approach that utilizes 2D foundation models to predict 3D correspondences. Specifically, ZeroReg adopts an object-to-point matching strategy, starting with object localization and semantic feature extraction from multi-view images using foundation models. In the object matching stage, semantic features help identify correspondences between objects across views. However, relying solely on semantic features can lead to ambiguity, especially in scenes with multiple instances of the same category. To address this, we construct scene graphs to capture spatial relationships among objects and apply a graph matching algorithm to these graphs to accurately identify matched objects. Finally, computing fine-grained point-level correspondences within matched object regions using algorithms like SuperGlue and LoFTR achieves robust point cloud registration. Evaluations on benchmarks such as 3DMatch, 3DLoMatch, and ScanNet demonstrate ZeroReg's competitive performance, highlighting its potential to advance point-cloud registration by integrating semantic features from foundation models.

The year 1953 is pivotal for computational physics: the first application of the Monte-Carlo method is published and calculations of the so-called Fermi-Pasta-Ulam-Tsingou experiment are started. It is the beginning of the massive use in the physical sciences of numerical methods implemented on electronic computers and a decisive step in the development of modern nonlinear dynamics. This will lead to an unpredictable development during the following 70 years. We briefly review the unfolding of these events and present some recent results that show how the issues raised are still relevant today, Comment: Invited article for Giornale di Fisica (English translation of the original Italian version)

Over the last decade, several regulatory bodies have started requiring the disclosure of non-financial information from publicly listed companies, in light of the investors' increasing attention to Environmental, Social, and Governance (ESG) issues. Publicly released information on sustainability practices is often disclosed in diverse, unstructured, and multi-modal documentation. This poses a challenge in efficiently gathering and aligning the data into a unified framework to derive insights related to Corporate Social Responsibility (CSR). Thus, using Information Extraction (IE) methods becomes an intuitive choice for delivering insightful and actionable data to stakeholders. In this study, we employ Large Language Models (LLMs), In-Context Learning, and the Retrieval-Augmented Generation (RAG) paradigm to extract structured insights related to ESG aspects from companies' sustainability reports. We then leverage graph-based representations to conduct statistical analyses concerning the extracted insights. These analyses revealed that ESG criteria cover a wide range of topics, exceeding 500, often beyond those considered in existing categorizations, and are addressed by companies through a variety of initiatives. Moreover, disclosure similarities emerged among companies from the same region or sector, validating ongoing hypotheses in the ESG literature. Lastly, by incorporating additional company attributes into our analyses, we investigated which factors impact the most on companies' ESG ratings, showing that ESG disclosure affects the obtained ratings more than other financial or company data.

Optimizing traffic dynamics in an evolving transportation landscape is crucial, particularly in scenarios where autonomous vehicles (AVs) with varying levels of autonomy coexist with human-driven cars. While optimizing Reinforcement Learning (RL) policies for such scenarios is becoming more and more common, little has been said about realistic evaluations of such trained policies. This paper presents an evaluation of the effects of AVs penetration among human drivers in a roundabout scenario, considering both quantitative and qualitative aspects. In particular, we learn a policy to minimize traffic jams (i.e., minimize the time to cross the scenario) and to minimize pollution in a roundabout in Milan, Italy. Through empirical analysis, we demonstrate that the presence of AVs} can reduce time and pollution levels. Furthermore, we qualitatively evaluate the learned policy using a cutting-edge cockpit to assess its performance in near-real-world conditions. To gauge the practicality and acceptability of the policy, we conduct evaluations with human participants using the simulator, focusing on a range of metrics like traffic smoothness and safety perception. In general, our findings show that human-driven vehicles benefit from optimizing AVs dynamics. Also, participants in the study highlight that the scenario with 80% AVs is perceived as safer than the scenario with 20%. The same result is obtained for traffic smoothness perception.

We seek to understand the composition and variability of fast (FSW) and slow Alfvenic solar wind (SASW) emerging from coronal holes (CH). We leverage an opportune conjunction between Solar Orbiter and Parker Solar Probe (PSP) during PSP Encounter 11 to include compositional diagnostics from the Solar Orbiter heavy ion sensor (HIS) as these variations provide crucial insights into the origin and nature of the solar wind. We use Potential Field Source Surface (PFSS) and Magnetohydrodynamic (MHD) models to connect the observed plasma at PSP and Solar Orbiter to its origin footpoint in the photosphere, and compare these results with the in situ measurements. A very clear signature of a heliospheric current sheet (HCS) crossing as evidenced by enhancements in low FIP elements, ion charge state ratios, proton density, low-Alfvenicity, and polarity estimates validates the combination of modeling, data, and mapping. We identify two FSW streams emerging from small equatorial coronal holes (CH) with low ion charge state ratios, low FIP bias, high-Alfvenicity, and low footpoint brightness, yet anomalously low alpha particle abundance for both streams. We identify high-Alfvenicity slow solar wind emerging from the over-expanded boundary of a CH having intermediate alpha abundance, high-Alfvenicity, and dips in ion charge state ratios corresponding to CH boundaries. Through this comprehensive analysis, we highlight the power of multi-instrument conjunction studies in assessing the sources of the solar wind., Comment: 23 pages, 10 figures

Malicious use of deepfakes leads to serious public concerns and reduces people's trust in digital media. Although effective deepfake detectors have been proposed, they are substantially vulnerable to adversarial attacks. To evaluate the detector's robustness, recent studies have explored various attacks. However, all existing attacks are limited to 2D image perturbations, which are hard to translate into real-world facial changes. In this paper, we propose adversarial head turn (AdvHeat), the first attempt at 3D adversarial face views against deepfake detectors, based on face view synthesis from a single-view fake image. Extensive experiments validate the vulnerability of various detectors to AdvHeat in realistic, black-box scenarios. For example, AdvHeat based on a simple random search yields a high attack success rate of 96.8% with 360 searching steps. When additional query access is allowed, we can further reduce the step budget to 50. Additional analyses demonstrate that AdvHeat is better than conventional attacks on both the cross-detector transferability and robustness to defenses. The adversarial images generated by AdvHeat are also shown to have natural looks. Our code, including that for generating a multi-view dataset consisting of 360 synthetic views for each of 1000 IDs from FaceForensics++, is available at https://github.com/twowwj/AdvHeaT.