119092 research outputs found
Sort by
ChartAttack: Testing the Vulnerability of LLMs to Malicious Prompting in Chart Generation
No full textMultimodal large language models (MLLMs) are increasingly used to automate chart generation from data tables, enabling efficient data analysis and reporting but also introducing new misuse risks. In this work, we introduce ChartAttack, a novel framework for evaluating how MLLMs can be misused to generate misleading charts at scale. ChartAttack injects misleaders into chart designs, aiming to induce incorrect interpretations of the underlying data. Furthermore, we create AttackViz, a chart question-answering (QA) dataset where each (chart specification, QA) pair is labeled with effective misleaders and their induced incorrect answers. Experiments in in-domain and cross-domain settings show that ChartAttack significantly degrades the QA performance of MLLM readers, reducing accuracy by an average of 19.6 points and 14.9 points, respectively. A human study further shows an average 20.2 point drop in accuracy for participants exposed to misleading charts generated by ChartAttack. Our findings highlight an urgent need for robustness and security considerations in the design, evaluation, and deployment of MLLM-based chart generation systems. We make our code and data publicly available
LLMs as Cultural Archives: Cultural Commonsense Knowledge Graph Extraction
No full textLarge language models (LLMs) encode rich cultural knowledge learned from diverse web-scale data, offering an unprecedented opportunity to model cultural commonsense at scale. Yet this knowledge remains mostly implicit and unstructured, limiting its interpretability and use. We present an iterative, prompt-based framework for constructing a Cultural Commonsense Knowledge Graph (CCKG) that treats LLMs as cultural archives, systematically eliciting culture-specific entities, relations, and practices and composing them into multi-step inferential chains across languages. We evaluate CCKG on five countries with human judgments of cultural relevance, correctness, and path coherence. We find that the cultural knowledge graphs are better realized in English, even when the target culture is non-English (e.g., Chinese, Indonesian, Arabic), indicating uneven cultural encoding in current LLMs. Augmenting smaller LLMs with CCKG improves performance on cultural reasoning and story generation, with the largest gains from English chains. Our results show both the promise and limits of LLMs as cultural technologies and that chain-structured cultural knowledge is a practical substrate for culturally grounded NLP
Anion exchange in LaBaInO3F2 : structural and optical effects of low-temperature topochemical modification
No full textWe present a comprehensive investigation of the low-temperature topochemical modification of the Ruddlesden–Popper-type (RP) indium oxyfluoride LaBaInO3F2 using sodium hydride (NaH). By varying NaH equivalents, we monitored phase evolution, anion exchange, and reduction pathways using combined structural, spectroscopic, elemental and the density functional theory (DFT) analysis. LaBaInO3F2 undergoes controlled anion-extraction, with hydride-for-fluoride substitution produces mixed-anion oxyfluoride-hydride phases, LaBaInO3F2–xHy, and NaF as a stable byproduct. At moderate xNaH contents (x ≤ 1), the RP-framework incorporates hydride with minimal decomposition, supported by reoxidation experiments and DFT-calculated reaction enthalpies that confirm oxidizable hydride and subtle modification of the indium coordination environment. In contrast, higher xNaH contents (x > 1) result in the formation of secondary phases, indicating progressive lattice degradation. X-ray photoemission spectroscopy (XPS) indicates the possibility of subtle reduction of indium, with redox changes secondary to anion exchange. The executed study revealed a pathway how to stabilize hydride ions next to In3+, which is known to be notoriously hard otherwise. Optical spectroscopy reveals bandgap narrowing and enhanced visible-light absorption in hydride-rich samples, however, DFT estimates that idealized hydride substitution widens the bandgap, consistent with the experimentally non-ideal reduction with secondary phases. Overall, these findings establish hydride-based topochemical strategies as a promising route for controlled anion exchange with implications for photocatalytic and energy applications
Exploring diverse solutions in network dismantling with generative model
No full textFinding an optimal subset of nodes to dismantle a complex network is a fundamental NP-hard problem. Existing algorithms typically yield only a single solution, limiting their practical use. To address this, we introduce Generative Flow Dismantling (GFD), a novel framework based on Generative Flow Networks (GFlowNets), which reformulates the dismantling task as a learnable, sequential sampling process. Trained on small synthetic networks, GFD generalizes to diverse real-world scenarios, efficiently producing a diverse set of high-quality dismantling strategies. Our data-driven analysis of these solutions reveals a key insight: the quality of a final dismantling set is more strongly correlated with the mean degree of the nodes in the initial decycling set than with its size alone. Leveraging this, we develop two enhanced variants, GFD-edges and GFD-nodes, that further improve solution quality and dismantling speed. By exploring the full landscape of diverse, high-quality solutions, our work establishes a new paradigm for network dismantling, offering a flexible and powerful framework with broad implications for understanding and enhancing network resilience
Application of advanced soil constitutive models to investigate complex soil-structure interaction issues in integral abutment bridges
No full textThe jointless design of integral bridges transfer deck expansions and contractions from ambient temperature fluctuations directly to the abutments. This cyclic behaviour leads to approach settlements and increases lateral earth pressures on the abutment. While numerical methods have been used to investigate this complex backfill-abutment interaction, most studies have relied on perfectly-plastic constitutive models for soils or have only validated passive pressure development over a limited number of cycles, neglecting settlement. Hence, previous research on numerical techniques have been unable to capture realistic service conditions, where integral abutments endure hundreds of cycles leading to densification and significant subsidence in the approach. This paper investigates the performance of three sophisticated constitutive models against the widely used Mohr-Coulomb (MC) model for simulating the backfill: SANISAND with Memory Surface and Semifluidized State (S-MSf), Hypoplasticity with Intergranular Strain (Hypo-IS) and Hypoplasticity with Intergranular Strain Anisotropy (Hypo-ISA). The backfill-abutment interaction is simulated using the Mohr-Coulomb friction (MCF) and Hypoplasticity with Intergranular Strain friction (Hypo-ISF) models. These models are carefully evaluated on simulations of centrifuge tests for integral abutments with different backfill relative densities, abutment heights and cyclic amplitudes. Results indicate that the MC and Hypo-IS constitutive models cannot capture the long-term backfill response. While integrated with the MCF contact, both S-MSf and Hypo-ISA deviated from experiments. The Hypo-ISA and Hypo-ISF pair estimated well the passive pressure development for the cases involving very large and very small abutment displacements. S-MSf demonstrated the best overall agreement with centrifuge data when paired with Hypo-ISF
Internal stress and dislocation-mediated phase structure and conduction mechanism in flash-sintered Na1/2Bi1/2TiO3-based ceramics
No full textNa1/2Bi1/2TiO 3 (NBT) is widely regarded as a multifunctional material, exhibiting dual capabilities as both an insulating piezoelectric material and an oxygen ion conductor. Historically, property tuning has predominantly relied on chemical doping to tailor composition and phase structure. Here,
0.94(Na0.5Bi0.5TiO3)–0.06BaTiO3 (NBT-6BT) ceramics were prepared by flash sintering. X-ray diffraction and scanning electron microscopy analyses indicate that the flash-sintered sample exhibit a heterogeneous grain morphology and features a high internal stress, which drives a rhombohedral-to-tetragonal transition and enhances ferroelectric behavior. Notably, transmission electron microscopy reveals dislocations within the ceramics. Furthermore, impedance spectroscopy demonstrates ionic
conductivity in undoped, flash-sintered NBT-6BT ceramics. Molecular dynamics (MD) simulations demonstrates that the enhanced ionic conductivity is associated with dislocation-induced pathways for oxygen diffusion. By showing that stress and dislocation engineering can effectively tune both structural and electrical properties without additional compositional modification, this study presents a compelling alternative to conventional chemical doping strategies for piezoelectric ceramic
Two-component anomalous Hall and Nernst effects in anisotropic Fe4–xGexN thin films
No full textA series of thin films Fe4–xGexN (x = 0 – 1) were fabricated onto MgO substrates by magnetron sputtering with the aim of studying the possible enhancement of the anomalous Nernst effect (ANE), envisaged based on density functional theory (DFT) calculations. The Nernst and Hall effects of the series were systematically analyzed, complemented with resistivity, magnetic, electron microscopy, and Mössbauer experiments, and DFT calculations including elastic properties. The Fe4N phase crystallizes in the cubic symmetry with Pm3̅m space group, whereas a small tetragonal distortion is realized in Fe4–xGexN films for x > 0.35. From the comparison of the experimental isomer shift with DFT calculations, we conclude that Ge occupies the 4b site in the tetragonal I4/mcm structure. The ferromagnetic Curie temperature decreases rapidly from ∼750 K for x = 0 to ∼100 K for x = 1. The tetragonal samples with x = 0.8 and 1 display two-component behavior in the Hall and Nernst effects hysteresis loops, which can be analyzed as a sum of positive and negative loops with different saturation fields. This unusual behavior is a product of a combination of several factors: (1) coexistence of two different crystallographic orientations in the tetragonal thin film, namely with the majority of c-axis and minority of a-axis normal to the film surface; (2) opposite sign of the anomalous Hall and Nernst effects for the direction of magnetization along the a- and c-axes revealed by DFT calculation; and (3) the magnetocrystalline anisotropy characterized by an easy ab-plane, which is responsible for the different saturation fields for a- and c-axes. The maximum ANE was determined to be 0.9 μV/K for x = 0 at room temperature and −0.85 μV/K for x = 1 at T = 50 K. The rapid increase of ANE of Fe3GeN from low temperatures indicates that, were it not for its low Curie temperature, it could surpass the ANE of Fe4N. This observation is consistent with our theoretical assumptions and motivates further research of doped Fe4N for which ANE enhancement is predicted by DFT calculations
Change point-based segmentation of machining cycles in electrical load curves of machine tools for data pre-processing
No full textMotivated by growing awareness of sustainability and rising costs, a detailed allocation of energy consumption attracts increasing attention. Nevertheless, the potential offered by the inherent information in electrical load profiles of production machines has not yet been fully utilised. This paper introduces an approach for segmenting the load profiles of machine tools for individual machining cycles, aiming to reduce the manual effort in data pre-processing of the time series, e.g. for clustering applications. The approach's applicability is examined on several machine tools. The performance of the approach is discussed using examples of load curves from machine tools
High‐pressure cell design for modulation excitation spectroscopy: application to mechanistic analysis of CO 2 hydrogenation to methanol
No full textModulation excitation spectroscopy (MES) is a powerful method to provide information on active species and sites in catalytic reactions. We present the design of a diffuse‐reflectance IR Fourier transform spectroscopy (DRIFTS) cell, which is suitable for MES but also applicable to other in situ and operando DRIFTS studies on catalytic materials. The cell is characterized by a low void volume to allow for sufficiently fast gas exchange during MES experiments, even at high gas pressures. The potential of the cell is illustrated for the mechanistic analysis of CO 2 hydrogenation over a ceria‐supported copper catalyst (Cu/CeO 2 ) at 10 bar using ME‐DRIFTS combined with phase‐sensitive detection (PSD). Using MES/PSD key intermediates of methanol formation could be identified, that is, carbonates and formates, as well as methoxy groups, which are immediate precursors of methanol. By comparison of the mechanistic behavior of Cu/CeO 2 with bare ceria, the crucial role of copper for the reaction toward methanol was demonstrated. The presented DRIFTS cell offers high versatility for steady‐state and transient spectroscopic analysis under in situ/operando conditions, providing detailed mechanistic information, including the identification of intermediates during surface reactions over catalytic materials, facilitating their rational design
