Erchin Serpedin

Can Polat, Erchin Serpedin, Mustafa Kurban et al.

Generative models for materials have achieved strong performance on periodic bulk crystals, yet their ability to generalize across scale transitions to finite nanostructures remains largely untested. We introduce Crystal-to-Nanoparticle (C2NP), a systematic benchmark for evaluating generative models when moving between infinite crystalline unit cells and finite nanoparticles, where surface effects and size-dependent distortions dominate. C2NP defines two complementary tasks: (i) generating nanoparticles of specified radii from periodic unit cells, testing whether models capture surface truncation and geometric constraints; and (ii) recovering bulk lattice parameters and space-group symmetry from finite particle configurations, assessing whether models can infer underlying crystallographic order despite surface perturbations. Using diverse materials as a structurally consistent testbed, we construct over 170,000 nanoparticle configurations by carving particles from supercells derived from DFT-relaxed crystal unit cells, and introduce size-based splits that separate interpolation from extrapolation regimes. Experiments with state-of-the-art approaches, including diffusion, flow-matching, and variational models, show that even when losses are low, models often fail geometrically under distribution shift, yielding large lattice-recovery errors and near-zero joint accuracy on structure and symmetry. Overall, our results suggest that current methods rely on template memorization rather than scalable physical generalization. C2NP offers a controlled, reproducible framework for diagnosing these failures, with immediate applications to nanoparticle catalyst design, nanostructured hydrides for hydrogen storage, and materials discovery. Dataset and code are available at https://github.com/KurbanIntelligenceLab/C2NP.

Can Polat, Erchin Serpedin, Mustafa Kurban et al.

Despite rapid advances in molecular and materials machine learning, most models still lack physical transferability: they fit correlations across whole molecules or crystals rather than learning the quantum interactions between atomic pairs. Yet bonding, charge redistribution, orbital hybridization, and electronic coupling all emerge from these two-body interactions that define local quantum fields in many-body systems. We introduce QuantumCanvas, a large-scale multimodal benchmark that treats two-body quantum systems as foundational units of matter. The dataset spans 2,850 element-element pairs, each annotated with 18 electronic, thermodynamic, and geometric properties and paired with ten-channel image representations derived from l- and m-resolved orbital densities, angular field transforms, co-occupancy maps, and charge-density projections. These physically grounded images encode spatial, angular, and electrostatic symmetries without explicit coordinates, providing an interpretable visual modality for quantum learning. Benchmarking eight architectures across 18 targets, we report mean absolute errors of 0.201 eV on energy gap using GATv2, 0.265 eV on HOMO and 0.274 eV on LUMO using EGNN. For energy-related quantities, DimeNet attains 2.27 eV total-energy MAE and 0.132 eV repulsive-energy MAE, while a multimodal fusion model achieves a 2.15 eV Mermin free-energy MAE. Pretraining on QuantumCanvas further improves convergence stability and generalization when fine-tuned on larger datasets such as QM9, MD17, and CrysMTM. By unifying orbital physics with vision-based representation learning, QuantumCanvas provides a principled and interpretable basis for learning transferable quantum interactions through coupled visual and numerical modalities. Dataset and model implementations are available at https://github.com/KurbanIntelligenceLab/QuantumCanvas.

Can Polat, Erchin Serpedin, Mustafa Kurban et al.

Every generative model for crystalline materials harbors a critical structure size beyond which its outputs quietly become unreliable -- we call this the extrapolation frontier. Despite its direct consequences for nanomaterial design, this frontier has never been systematically measured. We introduce RADII, a radius-resolved benchmark of ${\sim}$75,000 nanoparticle structures (55-11,298 atoms) that treats radius as a continuous scaling knob to trace generation quality from in-distribution to out-of-distribution regimes under leakage-free splits. RADII provides frontier-specific diagnostics: per-radius error profiles pinpoint each architecture's scaling ceiling, surface-interior decomposition tests whether failures originate at boundaries or in bulk, and cross-metric failure sequencing reveals which aspect of structural fidelity breaks first. Benchmarking five state-of-the-art architectures, we find that: (i) all models degrade by ${\sim}13\%$ in global positional error beyond training radii, yet local bond fidelity diverges wildly across architectures -- from near-zero to over $2\times$ collapse; (ii) no two architectures share the same failure sequence, revealing the frontier as a multi-dimensional surface shaped by model family; and (iii) well-behaved models obey a power-law scaling exponent $α\approx 1/3$ whose in-distribution fit accurately predicts out-of-distribution error, making their frontiers quantitatively forecastable. These findings establish output scale as a first-class evaluation axis for geometric generative models. The dataset and code are available at https://github.com/KurbanIntelligenceLab/RADII.

Can Polat, Erchin Serpedin, Mustafa Kurban et al.

Large language models are increasingly applied to materials science reasoning, yet their behavior under physically structured distribution shifts remains poorly understood. We introduce SCALAR (Structural Consistency And Logic Across Regimes), a benchmark for evaluating geometric scale generalization and its connection to structural hallucination, consistency, and reasoning in materials foundation models. Given canonical crystal representations, models must reason about derived nanoparticle structures obtained through supercell expansion and geometric truncation across length scales spanning a few atoms to over 18,000 atoms, totaling $\approx$100,000 structures from DFT-validated unit cells. SCALAR defines three tasks. (i) CIF to property prediction. (ii) A Chain-of-Thought variant with explicit physics-grounded reasoning. (iii) Inverse retrieval identifying crystals from candidates given target properties. Outputs are evaluated via structured metrics capturing numeric error, hallucination, cross-prompt consistency, monotonic reasoning, output validity, and retrieval regret. Experiments across diverse foundation models reveal large, model-dependent shifts under explicit reasoning, often reducing hallucination and error, but frequently destabilizing consistency or validity. These results demonstrate that geometric scale generalization cannot be inferred from accuracy alone. Supplementary materials are available at https://github.com/KurbanIntelligenceLab/SCALAR.

Samir Abdaljalil, Parichit Sharma, Erchin Serpedin et al.

Hallucinations in large language models remain a persistent challenge, particularly in multilingual and generative settings where factual consistency is difficult to maintain. While recent models show strong performance on English-centric benchmarks, their behavior across languages, tasks, and hallucination types is not yet well understood. In this work, we introduce Halluverse-M^3, a dataset designed to enable systematic analysis of hallucinations across multiple languages, multiple generation tasks, and multiple hallucination categories. Halluverse-M^3 covers four languages, English, Arabic, Hindi, and Turkish, and supports two generation tasks: question answering and dialogue summarization. The dataset explicitly distinguishes between entity-level, relation-level, and sentence-level hallucinations. Hallucinated outputs are constructed through a controlled editing process and validated by human annotators, ensuring clear alignment between original content and hallucinated generations. Using this dataset, we evaluate a diverse set of contemporary open-source and proprietary language models on fine-grained hallucination detection. Our results show that question answering is consistently easier than dialogue summarization, while sentence-level hallucinations remain challenging even for the strongest models. Performance is highest in English and degrades in lower-resource languages, with Hindi exhibiting the lowest detection accuracy. Overall, Halluverse-M^3 provides a realistic and challenging benchmark for studying hallucinations in multilingual, multi-task settings. We release the dataset to support future research on hallucination detection and mitigation\footnote{https://huggingface.co/datasets/sabdalja/HalluVerse-M3}.

Cihat Keçeci, Katherine R. Davis, Erchin Serpedin

Data analysis and monitoring on smart grids are jeopardized by attacks on cyber-physical systems. False data injection attack (FDIA) is one of the classes of those attacks that target the smart measurement devices by injecting malicious data. The employment of machine learning techniques in the detection and localization of FDIA is proven to provide effective results. Training of such models requires centralized processing of sensitive user data that may not be plausible in a practical scenario. By employing federated learning for the detection of FDIA attacks, it is possible to train a model for the detection and localization of the attacks while preserving the privacy of sensitive user data. However, federated learning introduces new problems such as the personalization of the detectors in each node. In this paper, we propose a federated learning-based scheme combined with a hybrid deep neural network architecture that exploits the local correlations between the connected power buses by employing graph neural networks as well as the temporal patterns in the data by using LSTM layers. The proposed mechanism offers flexible and efficient training of an FDIA detector in a distributed setup while preserving the privacy of the clients. We validate the proposed architecture by extensive simulations on the IEEE 57, 118, and 300 bus systems and real electricity load data.

Osman Boyaci, M. Rasoul Narimani, Katherine Davis et al.

This study employs Infinite Impulse Response (IIR) Graph Neural Networks (GNN) to efficiently model the inherent graph network structure of the smart grid data to address the cyberattack localization problem. First, we numerically analyze the empirical frequency response of the Finite Impulse Response (FIR) and IIR graph filters (GFs) to approximate an ideal spectral response. We show that, for the same filter order, IIR GFs provide a better approximation to the desired spectral response and they also present the same level of approximation to a lower order GF due to their rational type filter response. Second, we propose an IIR GNN model to efficiently predict the presence of cyberattacks at the bus level. Finally, we evaluate the model under various cyberattacks at both sample-wise (SW) and bus-wise (BW) level, and compare the results with the existing architectures. It is experimentally verified that the proposed model outperforms the state-of-the-art FIR GNN model by 9.2% and 14% in terms of SW and BW localization, respectively.

Abdullah Hayajneh, Mohammad Shaqfeh, Erchin Serpedin et al.

This paper presents a novel machine learning framework to consistently detect, localize and rate congenital cleft lip anomalies in human faces. The goal is to provide a universal, objective measure of facial differences and reconstructive surgical outcomes that matches human judgments. The proposed method employs the StyleGAN2 generative adversarial network with model adaptation to produce normalized transformations of cleft-affected faces in order to allow for subsequent measurement of deformity using a pixel-wise subtraction approach. The complete pipeline of the proposed framework consists of the following steps: image preprocessing, face normalization, color transformation, morphological erosion, heat-map generation and abnormality scoring. Heatmaps that finely discern anatomic anomalies are proposed by exploiting the features of the considered framework. The proposed framework is validated through computer simulations and surveys containing human ratings. The anomaly scores yielded by the proposed computer model correlate closely with the human ratings of facial differences, leading to 0.942 Pearson's r score.

Cihat Keçeci, Mohammad Shaqfeh, Fawaz Al-Qahtani et al.

This paper proposes using communication pipelining to enhance the wireless spectrum utilization efficiency and convergence speed of federated learning in mobile edge computing applications. Due to limited wireless sub-channels, a subset of the total clients is scheduled in each iteration of federated learning algorithms. On the other hand, the scheduled clients wait for the slowest client to finish its computation. We propose to first cluster the clients based on the time they need per iteration to compute the local gradients of the federated learning model. Then, we schedule a mixture of clients from all clusters to send their local updates in a pipelined manner. In this way, instead of just waiting for the slower clients to finish their computation, more clients can participate in each iteration. While the time duration of a single iteration does not change, the proposed method can significantly reduce the number of required iterations to achieve a target accuracy. We provide a generic formulation for optimal client clustering under different settings, and we analytically derive an efficient algorithm for obtaining the optimal solution. We also provide numerical results to demonstrate the gains of the proposed method for different datasets and deep learning architectures.

Cihat Keçeci, Mohammad Shaqfeh, Hayat Mbayed et al.

Federated learning enables many applications benefiting distributed and private datasets of a large number of potential data-holding clients. However, different clients usually have their own particular objectives in terms of the tasks to be learned from the data. So, supporting federated learning with meta-learning tools such as multi-task learning and transfer learning will help enlarge the set of potential applications of federated learning by letting clients of different but related tasks share task-agnostic models that can be then further updated and tailored by each individual client for its particular task. In a federated multi-task learning problem, the trained deep neural network model should be fine-tuned for the respective objective of each client while sharing some parameters for more generalizability. We propose to train a deep neural network model with more generalized layers closer to the input and more personalized layers to the output. We achieve that by introducing layer types such as pre-trained, common, task-specific, and personal layers. We provide simulation results to highlight particular scenarios in which meta-learning-based federated learning proves to be useful.

Abdullah Hayajneh, Erchin Serpedin, Mohammad Shaqfeh et al.

A major obstacle when attempting to train a machine learning system to evaluate facial clefts is the scarcity of large datasets of high-quality, ethics board-approved patient images. In response, we have built a deep learning-based cleft lip generator designed to produce an almost unlimited number of artificial images exhibiting high-fidelity facsimiles of cleft lip with wide variation. We undertook a transfer learning protocol testing different versions of StyleGAN-ADA (a generative adversarial network image generator incorporating adaptive data augmentation (ADA)) as the base model. Training images depicting a variety of cleft deformities were pre-processed to adjust for rotation, scaling, color adjustment and background blurring. The ADA modification of the primary algorithm permitted construction of our new generative model while requiring input of a relatively small number of training images. Adversarial training was carried out using 514 unique frontal photographs of cleft-affected faces to adapt a pre-trained model based on 70,000 normal faces. The Frechet Inception Distance (FID) was used to measure the similarity of the newly generated facial images to the cleft training dataset, while Perceptual Path Length (PPL) and the novel Divergence Index of Severity Histograms (DISH) measures were also used to assess the performance of the image generator that we dub CleftGAN. We found that StyleGAN3 with translation invariance (StyleGAN3-t) performed optimally as a base model. Generated images achieved a low FID reflecting a close similarity to our training input dataset of genuine cleft images. Low PPL and DISH measures reflected a smooth and semantically valid interpolation of images through the transfer learning process and a similar distribution of severity in the training and generated images, respectively.

Rasul Khanbayov, Mohamed Rayan Barhdadi, Erchin Serpedin et al.

Unsupervised physical parameter estimation from video lacks a common benchmark: existing methods evaluate on non-overlapping synthetic data, the sole real-world dataset is restricted to single-body systems, and no established protocol addresses governing-equation identification. This work introduces IRIS, a high-fidelity benchmark comprising 220 real-world videos captured at 4K resolution and 60\,fps, spanning both single- and multi-body dynamics with independently measured ground-truth parameters and uncertainty estimates. Each dynamical system is recorded under controlled laboratory conditions and paired with its governing equations, enabling principled evaluation. A standardized evaluation protocol is defined encompassing parameter accuracy, identifiability, extrapolation, robustness, and governing-equation selection. Multiple baselines are evaluated, including a multi-step physics loss formulation and four complementary equation-identification strategies (VLM temporal reasoning, describe-then-classify prompting, CNN-based classification, and path-based labelling), establishing reference performance across all IRIS scenarios and exposing systematic failure modes that motivate future research. The dataset, annotations, evaluation toolkit, and all baseline implementations are publicly released.

Mohamed Rayan Barhdadi, Samir Abdaljalil, Rasul Khanbayov et al.

Current 4D representations decouple geometry, motion, and semantics: reconstruction methods discard interpretable motion structure; language-grounded methods attach semantics after motion is learned, blind to how objects move; and motion-aware methods encode dynamics as opaque per-point residuals without object-level organization. We propose 4D Synchronized Fields, a 4D Gaussian representation that learns object-factored motion in-loop during reconstruction and synchronizes language to the resulting kinematics through a per-object conditioned field. Each Gaussian trajectory is decomposed into shared object motion plus an implicit residual, and a kinematic-conditioned ridge map predicts temporal semantic variation, yielding a single representation in which reconstruction, motion, and semantics are structurally coupled and enabling open-vocabulary temporal queries that retrieve both objects and moments. On HyperNeRF, 4D Synchronized Fields achieves 28.52 dB mean PSNR, the highest among all language-grounded and motion-aware baselines, within 1.5 dB of reconstruction-only methods. On targeted temporal-state retrieval, the kinematic-conditioned field attains 0.884 mean accuracy, 0.815 mean vIoU, and 0.733 mean tIoU, surpassing 4D LangSplat (0.620, 0.433, and 0.439 respectively) and LangSplat (0.415, 0.304, and 0.262). Ablation confirms that kinematic conditioning is the primary driver, accounting for +0.45 tIoU over a static-embedding-only baseline. 4D Synchronized Fields is the only method that jointly exposes interpretable motion primitives and temporally grounded language fields from a single trained representation. Code will be released.

Samir Abdaljalil, Hasan Kurban, Khalid Qaraqe et al.

Large language models (LLMs) have shown strong performance across natural language reasoning tasks, yet their reasoning processes remain brittle and difficult to interpret. Prompting techniques like Chain-of-Thought (CoT) enhance reliability by eliciting intermediate reasoning steps or aggregating multiple outputs. However, they lack mechanisms for enforcing logical structure and assessing internal coherence. We introduce Theorem-of-Thought (ToTh), a novel framework that models reasoning as collaboration among three parallel agents, each simulating a distinct mode of inference: abductive, deductive, and inductive. Each agent produces a reasoning trace, which is structured into a formal reasoning graph. To evaluate consistency, we apply Bayesian belief propagation guided by natural language inference (NLI), assigning confidence scores to each step. The most coherent graph is selected to derive the final answer. Experiments on symbolic (WebOfLies) and numerical (MultiArith) reasoning benchmarks show that ToTh consistently outperforms CoT, Self-Consistency, and CoT-Decoding across multiple LLMs, while producing interpretable and logically grounded reasoning chains. Our findings suggest a promising direction for building more robust and cognitively inspired LLM reasoning. The implementation is available at https://github.com/KurbanIntelligenceLab/theorem-of-thought.

Samir Abdaljalil, Erchin Serpedin, Hasan Kurban

Large language models are increasingly used to answer and verify scientific claims, yet existing evaluations typically assume that a model must always produce a definitive answer. In scientific settings, however, unsupported or uncertain conclusions can be more harmful than abstaining. We study this problem through an abstention-aware verification framework that decomposes scientific claims into minimal conditions, audits each condition against available evidence using natural language inference (NLI), and selectively decides whether to support, refute, or abstain. We evaluate this framework across two complementary scientific benchmarks: SciFact and PubMedQA, covering both closed-book and open-domain evidence settings. Experiments are conducted with six diverse language models, including encoder-decoder, open-weight chat models, and proprietary APIs. Across all benchmarks and models, we observe that raw accuracy varies only modestly across architectures, while abstention plays a critical role in controlling error. In particular, confidence-based abstention substantially reduces risk at moderate coverage levels, even when absolute accuracy improvements are limited. Our results suggest that in scientific reasoning tasks, the primary challenge is not selecting a single best model, but rather determining when available evidence is sufficient to justify an answer. This work highlights abstention-aware evaluation as a practical and model-agnostic lens for assessing scientific reliability, and provides a unified experimental basis for future work on selective reasoning in scientific domains. Code is available at https://github.com/sabdaljalil2000/ai4science .

Samir Abdaljalil, Erchin Serpedin, Khalid Qaraqe et al.

Large language models (LLMs) often generate reasoning traces that appear coherent but rest on unsupported assumptions, leading to hallucinated conclusions. Prior work mainly addresses factual hallucinations or relies on post-hoc verification, leaving reasoning-induced hallucinations largely unaddressed. We propose Audit-of-Understanding (AoU), a framework that constrains inference to validated premises through three phases: (1) decomposing a query into candidate assumptions, (2) auditing their support, and (3) conditioning inference only on the validated subset. Formally, AoU is \emph{posterior-constrained inference}, connecting to selective prediction and rejection learning. Our contributions are threefold: (i) theoretical guarantees under perfect validation, (ii) excess-risk bounds under imperfect audits, and (iii) tractability analysis. Empirically, AoU improves both accuracy and faithfulness on GSM8K, MultiArith, and SVAMP, achieving up to +30% gains on GSM8K, +45% on MultiArith, and consistent +20--28% improvements on SVAMP over Chain-of-Thought, Self-Consistency, and CoT-Decoding. Code is available at https://anonymous.4open.science/r/audit-of-understanding-E28B.

Samir Abdaljalil, Erchin Serpedin, Khalid Qaraqe et al.

Large language models (LLMs) are increasingly applied in multilingual contexts, yet their capacity for consistent, logically grounded alignment across languages remains underexplored. We present a controlled evaluation framework for multilingual natural language inference (NLI) that generates synthetic, logic-based premise-hypothesis pairs and translates them into a typologically diverse set of languages. This design enables precise control over semantic relations and allows testing in both monolingual and mixed-language (code-switched) conditions. Surprisingly, code-switching does not degrade, and can even improve, performance, suggesting that translation-induced lexical variation may serve as a regularization signal. We validate semantic preservation through embedding-based similarity analyses and cross-lingual alignment visualizations, confirming the fidelity of translated pairs. Our findings expose both the potential and the brittleness of current LLM cross-lingual reasoning, and identify code-switching as a promising lever for improving multilingual robustness. Code available at: https://github.com/KurbanIntelligenceLab/nli-stress-testing

Can Polat, Hasan Kurban, Erchin Serpedin et al.

Evaluating foundation models for crystallographic reasoning requires benchmarks that isolate generalization behavior while enforcing physical constraints. This work introduces a multiscale multicrystal dataset with two physically grounded evaluation protocols to stress-test multimodal generative models. The Spatial-Exclusion benchmark withholds all supercells of a given radius from a diverse dataset, enabling controlled assessments of spatial interpolation and extrapolation. The Compositional-Exclusion benchmark omits all samples of a specific chemical composition, probing generalization across stoichiometries. Nine vision--language foundation models are prompted with crystallographic images and textual context to generate structural annotations. Responses are evaluated via (i) relative errors in lattice parameters and density, (ii) a physics-consistency index penalizing volumetric violations, and (iii) a hallucination score capturing geometric outliers and invalid space-group predictions. These benchmarks establish a reproducible, physically informed framework for assessing generalization, consistency, and reliability in large-scale multimodal models. Dataset and code are available at https://github.com/KurbanIntelligenceLab/StressTestingMMFMinCR.

Can Polat, Erchin Serpedin, Mustafa Kurban et al.

Most materials science datasets are limited to atomic geometries (e.g., XYZ files), restricting their utility for multimodal learning and comprehensive data-centric analysis. These constraints have historically impeded the adoption of advanced machine learning techniques in the field. This work introduces MultiCrystalSpectrumSet (MCS-Set), a curated framework that expands materials datasets by integrating atomic structures with 2D projections and structured textual annotations, including lattice parameters and coordination metrics. MCS-Set enables two key tasks: (1) multimodal property and summary prediction, and (2) constrained crystal generation with partial cluster supervision. Leveraging a human-in-the-loop pipeline, MCS-Set combines domain expertise with standardized descriptors for high-quality annotation. Evaluations using state-of-the-art language and vision-language models reveal substantial modality-specific performance gaps and highlight the importance of annotation quality for generalization. MCS-Set offers a foundation for benchmarking multimodal models, advancing annotation practices, and promoting accessible, versatile materials science datasets. The dataset and implementations are available at https://github.com/KurbanIntelligenceLab/MultiCrystalSpectrumSet.

Yassine Habchi, Hamza Kheddar, Yassine Himeur et al.

The rapid advancement of deep learning (DL) has transformed healthcare, particularly in cancer detection and diagnosis. DL surpasses traditional machine learning and human accuracy, making it a critical tool for identifying diseases. Despite numerous reviews on DL in healthcare, a comprehensive analysis of its role in cancer detection remains limited. Existing studies focus on specific aspects, leaving gaps in understanding its broader impact. This paper addresses these gaps by reviewing advanced DL techniques, including transfer learning (TL), reinforcement learning (RL), federated learning (FL), Transformers, and large language models (LLMs). These approaches enhance accuracy, tackle data scarcity, and enable decentralized learning while maintaining data privacy. TL adapts pre-trained models to new datasets, improving performance with limited labeled data. RL optimizes diagnostic pathways and treatment strategies, while FL fosters collaborative model development without sharing sensitive data. Transformers and LLMs, traditionally used in natural language processing, are now applied to medical data for improved interpretability. Additionally, this review examines these techniques' efficiency in cancer diagnosis, addresses challenges like data imbalance, and proposes solutions. It serves as a resource for researchers and practitioners, providing insights into current trends and guiding future research in advanced DL for cancer detection.

Samir Abdaljalil, Hasan Kurban, Erchin Serpedin

Large Language Models (LLMs) are increasingly used in various contexts, yet remain prone to generating non-factual content, commonly referred to as "hallucinations". The literature categorizes hallucinations into several types, including entity-level, relation-level, and sentence-level hallucinations. However, existing hallucination datasets often fail to capture fine-grained hallucinations in multilingual settings. In this work, we introduce HalluVerse25, a multilingual LLM hallucination dataset that categorizes fine-grained hallucinations in English, Arabic, and Turkish. Our dataset construction pipeline uses an LLM to inject hallucinations into factual biographical sentences, followed by a rigorous human annotation process to ensure data quality. We evaluate several LLMs on HalluVerse25, providing valuable insights into how proprietary models perform in detecting LLM-generated hallucinations across different contexts.

Samir Abdaljalil, Hasan Kurban, Parichit Sharma et al.

Large language models (LLMs) are increasingly deployed across diverse domains, yet they are prone to generating factually incorrect outputs - commonly known as "hallucinations." Among existing mitigation strategies, uncertainty-based methods are particularly attractive due to their ease of implementation, independence from external data, and compatibility with standard LLMs. In this work, we introduce a novel and scalable uncertainty-based semantic clustering framework for automated hallucination detection. Our approach leverages sentence embeddings and hierarchical clustering alongside a newly proposed inconsistency measure, SINdex, to yield more homogeneous clusters and more accurate detection of hallucination phenomena across various LLMs. Evaluations on prominent open- and closed-book QA datasets demonstrate that our method achieves AUROC improvements of up to 9.3% over state-of-the-art techniques. Extensive ablation studies further validate the effectiveness of each component in our framework.

Samir Abdaljalil, Filippo Pallucchini, Andrea Seveso et al.

Despite the state-of-the-art performance of Large Language Models (LLMs), these models often suffer from hallucinations, which can undermine their performance in critical applications. In this work, we propose SAFE, a novel method for detecting and mitigating hallucinations by leveraging Sparse Autoencoders (SAEs). While hallucination detection techniques and SAEs have been explored independently, their synergistic application in a comprehensive system, particularly for hallucination-aware query enrichment, has not been fully investigated. To validate the effectiveness of SAFE, we evaluate it on two models with available SAEs across three diverse cross-domain datasets designed to assess hallucination problems. Empirical results demonstrate that SAFE consistently improves query generation accuracy and mitigates hallucinations across all datasets, achieving accuracy improvements of up to 29.45%.

Can Polat, Hasan Kurban, Erchin Serpedin et al.

Molecular graph neural networks (GNNs) often focus exclusively on XYZ-based geometric representations and thus overlook valuable chemical context available in public databases like PubChem. This work introduces a multimodal framework that integrates textual descriptors, such as IUPAC names, molecular formulas, physicochemical properties, and synonyms, alongside molecular graphs. A gated fusion mechanism balances geometric and textual features, allowing models to exploit complementary information. Experiments on benchmark datasets indicate that adding textual data yields notable improvements for certain electronic properties, while gains remain limited for others. Furthermore, the GNN architectures display similar performance patterns (improving and deteriorating on analogous targets), suggesting they learn comparable representations rather than distinctly different physical insights.

Mohamed Rayan Barhdadi, Mehmet Tuncel, Erchin Serpedin et al.

Current AI approaches to refugee integration optimize narrow objectives such as employment and fail to capture the cultural, emotional, and ethical dimensions critical for long-term success. We introduce EMPATHIA (Enriched Multimodal Pathways for Agentic Thinking in Humanitarian Immigrant Assistance), a multi-agent framework addressing the central Creative AI question: how do we preserve human dignity when machines participate in life-altering decisions? Grounded in Kegan's Constructive Developmental Theory, EMPATHIA decomposes integration into three modules: SEED (Socio-cultural Entry and Embedding Decision) for initial placement, RISE (Rapid Integration and Self-sufficiency Engine) for early independence, and THRIVE (Transcultural Harmony and Resilience through Integrated Values and Engagement) for sustained outcomes. SEED employs a selector-validator architecture with three specialized agents - emotional, cultural, and ethical - that deliberate transparently to produce interpretable recommendations. Experiments on the UN Kakuma dataset (15,026 individuals, 7,960 eligible adults 15+ per ILO/UNHCR standards) and implementation on 6,359 working-age refugees (15+) with 150+ socioeconomic variables achieved 87.4% validation convergence and explainable assessments across five host countries. EMPATHIA's weighted integration of cultural, emotional, and ethical factors balances competing value systems while supporting practitioner-AI collaboration. By augmenting rather than replacing human expertise, EMPATHIA provides a generalizable framework for AI-driven allocation tasks where multiple values must be reconciled.

Osman Boyaci, Mohammad Rasoul Narimani, Katherine Davis et al.

As a highly complex and integrated cyber-physical system, modern power grids are exposed to cyberattacks. False data injection attacks (FDIAs), specifically, represent a major class of cyber threats to smart grids by targeting the measurement data's integrity. Although various solutions have been proposed to detect those cyberattacks, the vast majority of the works have ignored the inherent graph structure of the power grid measurements and validated their detectors only for small test systems with less than a few hundred buses. To better exploit the spatial correlations of smart grid measurements, this paper proposes a deep learning model for cyberattack detection in large-scale AC power grids using Chebyshev Graph Convolutional Networks (CGCN). By reducing the complexity of spectral graph filters and making them localized, CGCN provides a fast and efficient convolution operation to model the graph structural smart grid data. We numerically verify that the proposed CGCN based detector surpasses the state-of-the-art model by 7.86 in detection rate and 9.67 in false alarm rate for a large-scale power grid with 2848 buses. It is notable that the proposed approach detects cyberattacks under 4 milliseconds for a 2848-bus system, which makes it a good candidate for real-time detection of cyberattacks in large systems.

Osman Boyaci, Mohammad Rasoul Narimani, Katherine Davis et al.

False data injection attacks (FDIA) are a main category of cyber-attacks threatening the security of power systems. Contrary to the detection of these attacks, less attention has been paid to identifying the attacked units of the grid. To this end, this work jointly studies detecting and localizing the stealth FDIA in power grids. Exploiting the inherent graph topology of power systems as well as the spatial correlations of measurement data, this paper proposes an approach based on the graph neural network (GNN) to identify the presence and location of the FDIA. The proposed approach leverages the auto-regressive moving average (ARMA) type graph filters (GFs) which can better adapt to sharp changes in the spectral domain due to their rational type filter composition compared to the polynomial type GFs such as Chebyshev. To the best of our knowledge, this is the first work based on GNN that automatically detects and localizes FDIA in power systems. Extensive simulations and visualizations show that the proposed approach outperforms the available methods in both detection and localization of FDIA for different IEEE test systems. Thus, the targeted areas can be identified and preventive actions can be taken before the attack impacts the grid.

Osman Boyaci, Amarachi Umunnakwe, Abhijeet Sahu et al.

False data injection attacks (FDIAs) represent a major class of attacks that aim to break the integrity of measurements by injecting false data into the smart metering devices in power grids. To the best of authors' knowledge, no study has attempted to design a detector that automatically models the underlying graph topology and spatially correlated measurement data of the smart grids to better detect cyber attacks. The contributions of this paper to detect and mitigate FDIAs are twofold. First, we present a generic, localized, and stealth (unobservable) attack generation methodology and publicly accessible datasets for researchers to develop and test their algorithms. Second, we propose a Graph Neural Network (GNN) based, scalable and real-time detector of FDIAs that efficiently combines model-driven and data-driven approaches by incorporating the inherent physical connections of modern AC power grids and exploiting the spatial correlations of the measurement. It is experimentally verified by comparing the proposed GNN based detector with the currently available FDIA detectors in the literature that our algorithm outperforms the best available solutions by 3.14%, 4.25%, and 4.41% in F1 score for standard IEEE testbeds with 14, 118, and 300 buses, respectively.

Mohamed Baza, Mahmoud Nabil, Muhammad Ismail et al.

Energy storage units (ESUs) enable several attractive features of modern smart grids such as enhanced grid resilience, effective demand response, and reduced bills. However, uncoordinated charging of ESUs stresses the power system and can lead to a blackout. On the other hand, existing charging coordination mechanisms suffer from several limitations. First, the need for a central charging coordinator (CC) presents a single point of failure that jeopardizes the effectiveness of the charging coordination. Second, a transparent charging coordination mechanism does not exist where users are not aware whether the CC is honest or not in coordination charging requests among them in a fair way. Third, existing mechanisms overlook the privacy concerns of the involved customers. To address these limitations, in this paper, we leverage the blockchain and smart contracts to build a decentralized charging coordination mechanism without the need for a centralized charging coordinator. First ESUs should use tokens for anonymously authenticate themselves to the blockchain. Then each ESU sends a charging request that contains its State-of-Charge (SoC), Time-to-complete-charge (TCC) and amount of required charging to the smart contract address on the blockchain. The smart contract will then run the charging coordination mechanism in a self-executed manner such that ESUs with the highest priorities are charged in the present time slot while charging requests of lower priority ESUs are deferred to future time slots. In this way, each ESU can make sure that charging schedules are computed correctly. Finally, we have implemented the proposed mechanism on the Ethereum test-bed blockchain, and our analysis shows that execution cost can be acceptable in terms of gas consumption while enabling decentralized charging coordination with increased transparency, reliability, and privacy preserving.

Mahmoud Nabil, Muhammad Ismail, Mohamed Mahmoud et al.

Modern smart grids rely on advanced metering infrastructure (AMI) networks for monitoring and billing purposes. However, such an approach suffers from electricity theft cyberattacks. Different from the existing research that utilizes shallow, static, and customer-specific-based electricity theft detectors, this paper proposes a generalized deep recurrent neural network (RNN)-based electricity theft detector that can effectively thwart these cyberattacks. The proposed model exploits the time series nature of the customers' electricity consumption to implement a gated recurrent unit (GRU)-RNN, hence, improving the detection performance. In addition, the proposed RNN-based detector adopts a random search analysis in its learning stage to appropriately fine-tune its hyper-parameters. Extensive test studies are carried out to investigate the detector's performance using publicly available real data of 107,200 energy consumption days from 200 customers. Simulation results demonstrate the superior performance of the proposed detector compared with state-of-the-art electricity theft detectors.