arXiv:2607.06025v1 Announce Type: new Abstract: Federated Learning (FL) offers a privacy-preserving framework for distributed machine learning, yet conventional centralized and hierarchical architectures present significant challenges in terms of scalability, resilience, and single points of failure, particularly in dynamic, infrastructure-less environments such as sensor networks. To address these limitations, we introduce FLAIR, a novel, fully decentralized FL protocol that integrates dynamic, resource-aware secure and self-organized clustering with in-cluster model training. FLAIR leverages a probabilistic, verifiable cluster-head election mechanism, which is enhanced to favor nodes with greater computational and communication capabilities, thereby ensuring both fairness and efficiency. Through comprehensive simulations in ns-3, we evaluate FLAIR against centralized, hierarchical, and gossip-based FL benchmarks across four demanding scenarios. The results demonstrate the superiority of our approach: in static 100-node networks, FLAIR achieves a final accuracy of approximately 0.91, outperforming all baselines. The protocol exhibits exceptional robustness, maintaining graceful degradation with accuracy above 0.85 even under 90% node failure rates. Furthermore, it shows strong resilience to mobility, with a performance loss of less than 2% compared to static deployments. In a realistic smart farming simulation, FLAIR's accuracy is within 0.2% of the centralized baseline, confirming its practical viability. These findings validate that FLAIR successfully combines the scalability of decentralized learning with the structural efficiency of clustering, presenting a robust and high performing solution for large-scale, heterogeneous IoT systems.
Science Journals
Splitting algorithms for paraxial and It\^o-Schr\"odinger models of wave propagation in random media
arXiv:2503.00633v2 Announce Type: replace Abstract: This paper introduces a full discretization procedure to solve wave beam propagation in random media modeled by a paraxial wave equation or an It\^o-Schr\"odinger stochastic partial differential equation. This method bears similarities with the phase screen method used routinely to solve such problems. The main axis of propagation is discretized by a centered splitting scheme with step $\Delta z$ while the transverse variables are treated by a spectral method after appropriate spatial truncation. The originality of our approach is its theoretical validity even when the typical wavelength $\theta$ of the propagating signal satisfies $\theta\ll\Delta z$. More precisely, we obtain a convergence of order $\Delta z$ in mean-square sense while the errors on statistical moments are of order $(\Delta z)^2$ as expected for standard centered splitting schemes. This is a surprising result as splitting schemes typically do not converge when $\Delta z$ is not the smallest scale of the problem. The analysis is based on equations satisfied by statistical moments in the It\^o-Schr\"odinger case and on integral (Duhamel) expansions for the paraxial model. Several numerical simulations illustrate and confirm the theoretical findings.
arXiv:2607.06421v1 Announce Type: new Abstract: Free-energy surfaces govern the populations of metastable states and the barriers that control transitions between them, making their direct optimization a central challenge in molecular and materials design. In this work, we introduce Gradient-Based Free Energy Surface Optimization (GB-FESO), an inverse design framework that uses a trained conditional diffusion model as a differentiable surrogate for the ensemble distribution. After training, the diffusion model is frozen, and the conditioning variables defining the system are optimized so that the generated ensemble reproduces a prescribed target free-energy surface. The optimization is carried out by backpropagating a distribution-level loss, based on kernel density estimates of the Kullback-Leibler divergence, through a deterministic diffusion sampling trajectory. We first validate GB-FESO on one-dimensional Gaussian ensembles, demonstrating that both continuous and relaxed discrete conditioning variables can be optimized to recover target distributions, including those outside the training domain. We then apply the method to a four-particle Lennard-Jones toy peptide exhibiting multiple metastable conformational states. In this more physically motivated setting, GB-FESO successfully optimizes the interaction parameters to reproduce target free-energy landscapes in the majority of test cases, with optimization performed either in the full internal-coordinate space or in a reduced collective-variable representation. These results establish GB-FESO as a promising first step toward an ensemble-level inverse design framework for molecular systems with prescribed thermodynamic and kinetic behavior.
arXiv:2607.06507v1 Announce Type: new Abstract: Multi-hop retrieval-augmented generation (RAG) acquires evidence sequentially, with each new document potentially revealing missing facts, bridge entities, query defects, or sufficient support for answering. Existing methods provide useful operations such as iterative retrieval, query reformulation, evidence critique, and sufficiency judging, but typically organize them within method-specific pipelines or predefined control topologies. This leaves underexplored how to learn a shared state-conditioned policy that chooses among currently valid evidence operations. We introduce DynaKRAG, which formulates multi-hop evidence acquisition as state-conditioned control over atomic evidence operations. At each step, a validity layer constructs the executable action set, and a learned controller selects the next operation. The resulting transition updates the evidence state and may enable new operations at subsequent steps. With Qwen2.5-7B-Instruct, DynaKRAG achieves F1 scores of 0.5998 on HotpotQA, 0.5340 on 2Wiki, and 0.3061 on MuSiQue, outperforming the strongest controlled baseline on all three benchmarks. Replacing the learned controller with a uniform-valid policy reduces F1 by 3.96--5.78 points, while removing sufficiency feedback hurts all three datasets. Controlled retrieval-cap experiments further show that additional retrieval is not uniformly beneficial. Together, these results demonstrate the benefit of coordinating retrieval, diagnosis, and gap-directed acquisition under an evolving evidence state.
arXiv:2508.01725v5 Announce Type: replace Abstract: Recent advances in continuous conditional generative modeling, including Continuous conditional Generative Adversarial Network (CcGAN) and Continuous Conditional Diffusion Model (CCDM), estimate high-dimensional data distributions conditioned on scalar regression labels such as angles, ages, or temperatures. However, fixed-size vicinal training in CcGAN can be sensitive to non-uniform label densities, whereas CCDM relies on computationally expensive iterative sampling. To address these issues, we propose CcGAN-AVAR, an imbalance-aware extension of CcGAN that combines soft/hybrid adaptive vicinity with auxiliary discriminator-guided regularization. The adaptive vicinity constructs a label-dependent local radius according to the available samples around each target condition, and the multi-task discriminator supplies both a regression signal for label consistency and a density-ratio-estimation signal for distribution matching. We further provide a theoretical interpretation characterizing how adaptive vicinal weighting affects the local bias-variance behavior of the discriminator target, how hybrid truncation reduces objective-level cross-condition mixing, and how the density-ratio-based generator penalty approximates a Pearson Chi-square discrepancy up to the estimation error of the density-ratio branch. Extensive experiments on four datasets, including the newly constructed imbalanced RC-49-I, covering resolutions from 64x64 to 256x256 across eleven settings, demonstrate that CcGAN-AVAR obtains strong generation quality and label consistency while preserving the one-step sampling efficiency of GANs, achieving 300x--2000x faster inference than CCDM.
arXiv:2607.05427v1 Announce Type: new Abstract: In this research, I analyze the gender dynamics of the global citation elite using annual top 2% Stanford/Elsevier lists for 2019-2024. My database includes 1.22 million person-year observations (N=1,221,363), which corresponds to 465,707 unique scientists and scholars from more than 150 countries. I move away from static representations of women in the citation elite toward analyses of entry, exit, and permanent membership in the durable core of this elite. The share of women in the annual citation elite increased from 18.39% in 2019 to 20.98% in 2024. However, women are more strongly represented among first-observed entrants than among continuing members, and their share decreases with the persistence of their presence among the citation elite expressed in years: from 22.19% among single-year members to 17.84% among scientists and scholars present in all six annual lists. Women are generally located closer to the lower boundary of the elite in terms of the citation index deciles - and men are closer to top deciles. My logistic regression models estimate a lower probability of women s membership in the durable core of the citation elite (odds ratio estimate OR=0.69). Women are also more weakly represented in the all-career elite than in the annual elite (15.87% vs. 20.98%). I draw conclusions about gender dynamics within the global citation elite and gender inequalities in science more generally.
arXiv:2607.05428v1 Announce Type: new Abstract: We present Charlie, an on-premise multi-agent Retrieval-Augmented Generation (RAG) system for structured evidential processing in digital forensic environments. Contemporary forensic workflows must handle large volumes of heterogeneous and unstructured documents under strict requirements of traceability, confidentiality, and legal compliance. Charlie addresses this challenge through a controlled agent architecture that combines local retrieval, task decomposition, structured memory, and verification mechanisms. Unlike cloud-based systems, it operates entirely within institutional infrastructure, preserving data sovereignty and evidential integrity. We describe the systems architecture, including its transition from classical RAG to agent-based orchestration, and demonstrate its application in real-world forensic scenarios. Case studies show that Charlie enables scalable multi-document data extraction and supports longitudinal forensic intelligence generation while maintaining traceability and auditability. Our results indicate that agent-orchestrated, on-premise RAG architectures can effectively support evidential workflows without compromising legal and institutional constraints. Charlie provides a practical and reproducible blueprint for deploying AI systems in high-stakes forensic environments. This manuscript is an archival version of a paper presented at the RELAF 2026 Workshop.
arXiv:2607.05993v1 Announce Type: new Abstract: Modern data centers increasingly rely on large-scale GPU clusters and on-site renewable energy resources, resulting in a tightly coupled cyber-physical system between computing workloads and power-electronic-dominated grids. In this paper, we reveal Bit2Watt, a previously unexplored vulnerability in which an adversary manipulates GPU workloads to induce controlled, high-frequency power modulations that destabilize local power infrastructure and propagate back to disrupt computing services. Unlike traditional attacks that compromise grid-side devices or communication channels, Bit2Watt operates entirely within the cyber layer as a legal tenant, which could amplify fluctuations, harmonic distortion, and damping degradation, particularly in high-DER-penetration scenarios. This risk is difficult to detect under routine cloud- and facility-side monitoring because it exploits legitimate workload execution paths and concentrates much of its distinctive behavior in high-frequency components that are weakly captured by common telemetry. We validate Bit2Watt through impedance-based analysis, power system simulations, and real-world experiments on GPUs and grid-connected PV inverters. Under the synchronized worst-case aggregation model studied in the paper, manipulating 1,000 GPUs in a 1-MW local power system with 90% DERs raises current THD to 46.8% and results in a damping ratio of -0.27. We further show that the resulting power-quality degradation can stress data-center power-delivery equipment, trigger protection mechanisms, and, in extreme simulated cases, induce cascading failures in transmission-scale systems. In addition, we analyze a plausible Watt2Bit feedback path, including denial-of-service risks and covert information exfiltration via EMI side channels. This work highlights the urgent need for cross-layer defenses that jointly consider workload scheduling and power electronics.
arXiv:2607.05429v1 Announce Type: new Abstract: Sixth-Generation (6G) networks will require massive connectivity, ultra-low latency, and robust security, making reliable Active User Detection (AUD) essential for interference control and physical layer protection. This letter proposes a Grover-based physical layer security (PLS) framework for a code-domain non-orthogonal multiple access (CD-NOMA) network, where the base station employs artificial-noise (AN)-assisted beamforming and identifies the active set via Grover's quantum search algorithm. We consider two threat models: passive eavesdroppers formed by detected inactive users, and active eavesdroppers selected as the top f% most frequent transmitters among detected active users. By aligning beams and AN with the Grover-based AUD output, the proposed scheme enlarges the main-wiretap rate gap and significantly improves the average secrecy rate compared with compressive sensing and classical correlation receiver baselines, while approaching maximum-likelihood detection performance with a quadratic reduction in search complexity. The impact of the information/AN power split, the base station transmit power, and the fraction of highly active users treated as eavesdroppers on secrecy is characterized through numerical simulations, and design insights are extracted for 6G PLS under both passive and active eavesdropping.
arXiv:2607.05434v1 Announce Type: new Abstract: Artificial Intelligence (AI) models, at their core, apply general learnings from broad datasets to individual circumstances using probabilistic behaviour. This inductive approach stands in contrast to deductive reasoning approaches which seek to prove conclusions from their premises. However, research has shown that deductive reasoning with AI models is a challenging problem and in the real-world it may not always be feasible. An alternative way forward is to leverage abductive reasoning, seeking to corroborate the output of multiple approaches to identify the most likely conclusion from the factual matrix. We apply this to synthetic media detection in forensic settings, and find we are able to disproportionately lower the risk of false positives to true positive recall. We also provide the first empirical evaluation of OpenAI's rollout of SynthID on synthetic images and evaluate how complementary different synthetic media detection approaches are.
arXiv:2607.05439v1 Announce Type: new Abstract: Many all-atom generative protein models can in principle design large multimeric complexes by jointly modelling all chains, but their quadratic token- and atom-pair representations quickly exceed single-GPU memory as the number of chains and residues modelled grows. We introduce Design-CP, two context-parallel (CP) inference strategies for RFdiffusion 3 (1D row-sharding and 2D grid sharding with ring attention) that distribute the quadratic activations across a multi-GPU mesh while preserving pretrained weights. We characterise their scaling when sampling icosahedral assemblies, showing that the maximum feasible asymmetric subunit (ASU) size grows with the expected square-root trend in GPU count and that 2D sharding achieves better wall-clock scaling. Moreover, we show how strong point-group symmetry constraints make CP usable out of the box for end-to-end, all-atom design of icosahedral nanoparticles, yielding favourable in silico structural and interface metrics. Finally, we demonstrate octahedral nanoparticle design on a small cluster of workstation-grade 16GB GPUs, illustrating how Design-CP can be a practical path towards democratising large-assembly protein design.
arXiv:2607.05724v1 Announce Type: new Abstract: Quantum convolutional neural networks (QCNNs) combine the power of quantum computing and classical CNN for computational speedup in classification tasks. However, noise levels on state-of-the-art quantum devices remain too high for practical QCNN execution. In addition, despite the reliable surface code providing a method for error rates below a threshold value, they have a prohibitively large qubit cost. Recently introduced bivariate bicycle (BB) codes are of particular interest for their high error threshold, constant encoding rate, and linear code distance. Through simulation with realistic hardware noise sources, we demonstrate that a 4-qubit unprotected QCNN fails to converge and exhibits a worse learning rate compared to numerical simulations. Addressing both limitations, we propose a distance-4 BB quantum error-correction (QEC) technique for QCNNs. In doing so, we validate that our low-overhead QEC technique for QCNNS represents a step toward practical QCNNs.
arXiv:2607.05898v1 Announce Type: new Abstract: Evaluating whether unlearning algorithms truly remove training data influence remains an open challenge. We propose a practical auditor that computes data-dependent lower bounds on the unlearning parameter $\varepsilon$ using membership inference attacks. Evaluating multiple unlearning algorithms, we find a sharp separation: algorithms with rigorous guarantees, such as model clipping and rewind-to-delete, achieve very small $\varepsilon$ bounds that do not falsify their unlearning guarantees, whereas empirical methods such as Hessian-based unlearning, interleaved ascent-descent, ascent on the forget set, and fine-tuning on the retain set exhibit large bounds, indicating poor unlearning. Our auditor provides a practical tool for empirically falsifying unlearning claims through a hypothesis-testing framework, and we validate it on CIFAR-100 and Shakespeare text.
arXiv:2607.05442v1 Announce Type: new Abstract: Responsible vulnerability disclosure can secure the defender's head start by controlling when a vulnerability becomes public. However, this status quo is now challenged by increases in capability of AI models, which benefits both defenders and adversaries. When both sides draw their capability from the same AI model, the defender's head start depends on the lab's decision to release the model, and the question becomes not whether to release but how. Existing safety frameworks govern only the deploy-or-withhold threshold and leave the timing of release unmodeled. We cast this decision as a bilevel Stackelberg game in which a lab commits to a window that sets each side's capability over time in a downstream contest between defender and adversary. Defender welfare turns on the capability gap, not the shared level. Handing one model to both sides can trap the defender in a Red Queen's race, whereas a pre-release to the defender alone creates a protective gap, and the lab's optimal window balances this welfare gain against the opportunity cost of delaying release. For dual-use models, the lever is the sequencing of access, not the deployment threshold.
arXiv:2607.05446v1 Announce Type: new Abstract: We introduce an interpolatory subdivision scheme for control polygons that take values in the three-dimensional Heisenberg group, the simplest noncommutative model geometry. The scheme keeps existing points at every refinement step and inserts new ones by a coordinate rule whose central correction comes from the group law. The two horizontal coordinates are refined by the classical four-point scheme of Dyn, Gregory and Levin, while the central coordinate acquires a closed-form correction built from a signed area of neighbouring horizontal data. Our main finding concerns the regularity of the limit curve. The horizontal part is exactly the classical four-point limit and inherits its smoothness. The central part behaves very differently. We prove that it converges to a continuous limit that belongs to the Zygmund class, with a logarithmic modulus of continuity. Under an explicit and verifiable condition on the central forcing, this logarithmic bound is sharp, because the scaled first differences then grow linearly with the refinement level, and the limit fails to be continuously differentiable. The effect is confirmed numerically. The correction is harmless at any single refinement step, but its repeated injection at every scale is what impacts smoothness. The example serves as a caution for nonlinear and group-valued subdivision, where a geometrically natural correction can impact regularity.
arXiv:2607.05952v1 Announce Type: new Abstract: While signed social recommendation has shown great potential by modeling both trust and distrust relations, its effectiveness is often hindered by structural noise and data sparsity. In this work, we first identify a fundamental inconsistency across the structural, propagation, and semantic layers of existing models, which leads to biased representations learned from sparse or noisy datasets. Furthermore, we observe that most existing methods treat the observed graph as fixed, failing to bridge the gap between noisy topologies and reliable social semantics. To address these issues, we propose a unified framework named SSC-Loop that treats signed social recommendation as the maximization of structural consistency. SSC-Loop includes three dedicated modules: ESA-DA for structural consistency, a P/N/O propagation mechanism for propagation consistency, and a contrastive learning objective for semantic consistency. Experiments on Epinions demonstrate that SSC-Loop achieves strong performance on explicit signed social rating prediction, while auxiliary results on Slashdot under a derived link-existence setting further suggest its ability to exploit signed social structures. Source code is available at https://github.com/Refrainwww/SSC-Loop.
arXiv:2607.05992v1 Announce Type: new Abstract: Mathematical reasoning has become a central task for evaluating and tuning reasoning Large Language Models (LLMs), yet existing benchmarks remain heavily biased toward high-resource languages, with English and Chinese dominating both pre-training corpora and evaluation suites. The recently released PolyMath (Wang et al., 2025) dataset represents a significant step forward, yet its coverage is still limited to 18 only high-resource languages. To address this gap, we introduce PluraMath, an extension of PolyMath to 18 additional {underrepresented languages spanning 6 language families -- ranging from mid-resource to extreme low-resource settings. We constructed the dataset through a human-curated pipeline, where native speakers thoroughly validated pre-computed translations. Using PluraMath, we then benchmark 27 reasoning LLMs across four model scales -- small, mid-size, large, and closed-source ensembles -- probing the multilingual mathematical reasoning capabilities of state-of-the-art models under diverse linguistic conditions. Our fine-grained analysis confirms a persistent gap in mathematical reasoning performance between high-resource and underrepresented languages, with stronger results largely associated with better instruction-following ability. We fully open-source our dataset, data acquisition pipeline, and evaluation framework, with the goal of lowering the barrier to multilingual benchmark development for underrepresented communities.
arXiv:2607.06017v1 Announce Type: new Abstract: We study a human-AI service system in which tasks arrive sequentially and are processed through a two-stage architecture: an automated chatbot followed, when necessary, by a human agent. We consider $T$ sequentially arriving tasks, each belonging to one of $K$ heterogeneous types. For each task the decision maker chooses how many resources to allocate to the chatbot, whose type-dependent success probabilities are initially unknown. Tasks not resolved by the chatbot enter type-dependent human-service queues, where they are processed by a human agent with unknown service rates. This model captures a central tradeoff in hybrid service systems: relying more on automation reduces human congestion but increases chatbot costs, while insufficient automation may overload the human agent. We propose the UCB-DPP policy, which combines Upper Confidence Bounds with Drift-Plus-Penalty control to learn the unknown parameters of the system while making queue-aware decisions. We prove that UCB-DPP achieves regret $\widetilde{\mathcal{O}}(K\sqrt{T})$ and guarantees mean-rate stability of the human-service queues. Simulations on synthetic instances show that the proposed policy outperforms natural baselines.
arXiv:2607.05452v1 Announce Type: new Abstract: Time series forecasters that use exogenous covariates are fragile in deployment: when those covariates are noised, temporally misaligned, or missing, strong exogenous-fusion and exogenous-adapted models can degrade far above the endogenous-only floor. We study whether such robustness requires specialized architectures, or whether it can be obtained through a simple training intervention. We propose exogenous dropout, a model-agnostic method that randomly zeros whole exogenous channels during training. Across electricity-price forecasting, reservoir hydrology, and meteorology, exogenous dropout substantially improves robustness under Gaussian noise, temporal misalignment, and fully missing channels, while preserving clean accuracy. Applied to a dual-correlation network, it yields the most robust model in our experiments, outperforming a deliberately strong bounded architectural foil, BoundEx, which combines a learnable gate, a fallback residual to the endogenous backbone, and per-channel exogenous FiLM modulation. Architecture-by-dropout ablations, gate-behavior diagnostics, and a representation-level bound show that explicit architectural boundedness is not necessary for this robustness: an unbounded model trained with exogenous dropout is more robust than the bounded model in every domain. We release a corruption-robustness benchmark and recommend exogenous dropout as a simple, strong baseline for future work on time series forecasting with covariates.
arXiv:2607.05804v1 Announce Type: new Abstract: On-policy distillation (OPD) trains a student policy by matching a stronger teacher on the student's own trajectories, offering a promising framework for language agent training. However, its application to long-horizon agentic tasks remains insufficiently explored. We identify two key inefficiencies in vanilla agent OPD: (1) full-horizon rollouts often waste wall-clock resources on tail turns that provide weak and noisy KL supervision, and (2) trajectory-level KL objectives concentrate most of the loss on shallow tokens, leaving deeper decision turns under-trained once initial behaviors are aligned. To address these challenges, we propose TurnOPD, a turn-level budgeting strategy for efficient on-policy distillation of long-horizon agents. TurnOPD consists of two budget controllers: adaptive rollout-depth budgeting, which uses probe-based turn statistics to determine rollout length, and progressive turn-normalized loss budgeting, which gradually shifts KL weighting from token-level to turn-balanced supervision. Experiments on ALFWorld, WebShop, and Multi-Hop Search with task-specialized teacher models show that TurnOPD achieves superior validation accuracy under equal wall-clock training budgets and advances the accuracy--time frontier beyond vanilla OPD.
arXiv:2606.20416v2 Announce Type: replace Abstract: Diffusion models rely heavily on explicit timestep embeddings to modulate the denoising process across various noise scales. In this work, we challenge the necessity of these temporal signals by analyzing their impact on U-Net and Diffusion Transformer architectures. Beyond empirical evidence, we provide a theoretical framework demonstrating that, under certain conditions, the global minimizer of the diffusion training objective can be achieved without explicit timestep conditioning. Our findings reveal a surprising robustness when timestep embeddings are completely removed. Extensive ablation studies on the CelebA and CIFAR-10 datasets show that these time-agnostic models can maintain high structural fidelity and even surpass their conditioned counterparts in competitive metrics, including FID, precision, and recall. Our analysis suggests these architectures can implicitly infer noise scales from the corrupted input under specific assumptions, rendering explicit temporal conditioning redundant. This study challenges long-standing temporal conditioning paradigms and paves the way for more efficient and structurally focused generative architectures.
arXiv:2606.22511v2 Announce Type: replace Abstract: In open-ended generation, LLMs frequently fall into the "likelihood trap", marked by repetitive degeneration and vocabulary dullness, creating a discrepancy between machine-generated and human-written text. While post-hoc tail truncation (e.g., Top-$p$, Min-$p$) avoids sampling from the unreliable tail, it can over-sample from the uncalibrated head and misalign generation with human lexical preferences; fixed scalar repetition penalties likewise ignore variation in logit scale across inference steps, potentially disrupting semantic coherence. To address both limitations, we propose Variance-Calibrated Modulation (VCM), a training-free pre-decoding intervention that reshapes the probability distribution before truncation through two dynamic mechanisms: (1) Contextual Searchlight via PMI, which suppresses global stopwords while elevating context-evoked tokens, and (2) Adaptive Self-Debiasing, which uses real-time logit standard deviation for scale-invariant penalization. Across open-ended generation, factual QA, and mathematical reasoning, VCM consistently mitigates the likelihood trap. With negligible computational overhead, VCM integrates with existing decoding strategies, improving diversity, coherence, and, particularly at higher decoding temperatures, reasoning accuracy.
arXiv:2607.05475v1 Announce Type: new Abstract: Deploying Large Language Models (LLMs) on mobile devices enhances privacy and reduces latency, but is severely bottlenecked by hardware inefficiency. We present the first comprehensive, cross-layer measurement study of mobile LLM inference, uniquely spanning five mainstream frameworks (e.g., llama.cpp, GENIE) and three hardware backends (CPU, GPU, NPU). To enable this analysis, we develop PowerBench, a fine-grained profiling tool that provides the first backend-specific energy attribution, moving beyond traditional device-level measurements. Our study yields three critical insights: (1) Framework-induced performance gaps are substantially amplified on NPUs, reaching up to 10x using custom operators due to divergent offloading and quantization strategies. (2) We identify a distinct phase split where NPUs excel at compute-bound prefilling, while CPUs outperform all other backends in memory-bound decoding. This is driven by the NPU's preference for large, fixed-shape workloads, which conflicts with the small-kernel, dynamic nature of decoding. (3) Backend-specific profiling uncovers substantial scheduling headroom missed by prior work. Suboptimal thread configurations, uncoordinated NPU sleep latencies, and CPU polling intervals result in up to 40% energy waste. Leveraging these findings, we present an energy-oriented best-practice configuration for mobile LLM inference. We estimate that this configuration could reduce energy consumption by up to 54.8% on the NPU backend across three datasets.
arXiv:2607.06466v1 Announce Type: new Abstract: This paper addresses the remote verification of the authenticity of Optically Variable Devices (commonly known as holograms) on identity documents. Typically placed over the cardholder's photo, these devices provide strong and easily verifiable security for human inspection but pose challenges for automated verification. Existing approaches easily cover static frauds (e.g. paper photocopy) and can be evaluated for such, but their capacity to detect real, dynamic fraud cases (e.g. handcrafted hologram) has not been evaluated to date because of the lack of public datasets. Furthermore, they are usually trained to detect known attack types, and few of them can generalize to new, unseen attacks. This work features three contributions to address these limitations: 1) a new public dataset, MIDV-DynAttack, which extends the existing MIDV-Holo dataset with realistic, static and dynamic attacks against identity document specimens, tripling the number of attack samples compared to the original dataset, 2) a novel verification method which can assess the authenticity of a specific hologram thanks to the analysis of its dynamic behavior and appearance, can be trained without dynamic attack samples, and exhibits new state-of-the-art performance, 3) a benchmark of existing approaches which follows a clear evaluation protocol and emphasizes the inability of other approaches to deal with dynamic attacks, as well as new challenging attacks to deal with. Code and dataset are publicly available at https://github.com/EPITAResearchLab/pouliquen.25.icdar.
arXiv:2607.05479v1 Announce Type: new Abstract: Generative AI (GenAI) systems store and process client data in three distinct ways: in the model's parameters through training and memorisation, in the context window during a live session, and in knowledge databases for retrieval-augmented generation (RAG). Each mode creates different and often counter-intuitive risks to confidentiality and legal professional privilege, and each calls for specific governance responses. Drawing on the first English and American decisions to address privilege and generative AI, UK and Munir v Secretary of State for the Home Department and United States v Heppner, on the orthodox privilege authorities against which those decisions must be read, and on recent computer science research, we explain the three modes of data storage and processing in terms accessible to practitioners and analyse the legal consequences of each. We then situate the analysis within the regulatory framework governing solicitors in England and Wales and within the ordinary principles of professional negligence, arguing that the standard of effective information governance (and with it the benchmark against which negligence and misconduct will be measured) is changing. Although we write primarily for SRA-regulated practitioners, our data-governance analysis is framed to extend to any jurisdiction in which the protection of privilege or professional secrecy depends on demonstrable confidentiality. The ultimate aim of this article is to help legal services professionals understand salient data leakage risks in GenAI systems and thereby facilitate a more responsible deployment of GenAI on client data and other sensitive material.