Papers

Blockchain-enabled Internet of Things (IoT) systems integrate smart contracts with embedded devices to support decentralized device management and access control. Their security therefore depends jointly on the logic of on-chain contracts and off-chain device firmware. Logic flaws in either layer can violate the same system invariants, such as unauthorized access, improper state changes, or unguarded privileged operations. Existing approaches rely on contract analysis, firmware analysis, and graph-based vulnerability detection. However, these methods typically focus on a single layer or artifact and often depend on predefined vulnerability patterns, emulation fidelity, or homogeneous representations that obscure security-relevant component roles. They also lack a unified architecture that supports different security tasks while remaining deployable on resource-constrained gateways. To address these limitations, we extend MA-HGAT into a cross-layer multi-agent heterogeneous graph attention framework that models contracts, firmware artifacts, device fleets, and transaction streams with a unified four-role, nine-relation schema. Role-aligned agents exchange heterogeneous evidence through cross-attention, while graph-, link-, and node-level heads support multiple detection tasks and a role-based gateway–cloud partition enables lightweight edge inference. MA-HGAT thus provides a unified and deployable framework for detecting logic vulnerabilities across the contract and device layers of blockchain-enabled IoT systems.

We investigate how a dynamically responsive gas disk changes the influence of self-interacting dark matter (SIDM) on stellar-bar formation. In an isolated Milky-Way-mass galaxy with fixed total baryonic mass, the gas-rich collisionless (CDM) model develops a persistent bar at $1.96\,\Gyr$, whereas the matched SIDM model with $\sigmam=1\,\mathrm{cm^2\,g^{-1}}$ does so at $2.69\,\Gyr$. The final stellar $m=2$ amplitude is 0.437 in CDM and 0.340 in SIDM, a 22 per cent suppression; the corresponding gas-free controls remain weak over the same interval. An independently sampled gas-rich pair preserves the direction of both differences. Direct component-force measurements show that the SIDM stellar disk loses $2.82\times10^{11}\,\Msun\,\kpc\,\kms$ less angular momentum than its CDM counterpart. The integrated torque contrast is dominated by the dark halo (approximately 74 per cent), with a substantial gas contribution (26 per cent), and is redistributed in radius and time rather than being a uniform reduction of halo coupling. Conservation, force-accuracy, aperture, threshold, and independent-sampling checks support the ordering as a dynamical effect within this model. The result demonstrates that a live gas component can change both the sign and timing of the early SIDM impact inferred from collisionless disks, providing a mechanism-based constraint for future hydrodynamic SIDM studies.

The classical perceptron cycling theorem of Block and Levin \cite{BlockLevin1970} bounds correction sequences whose selected updates come from a finite set and have nonpositive inner product with the current state. We prove a robust variant for additive trajectories $z_{k+1}=z_k+u_k$, for all integers $k\geq0$, with increments in a finite set $U\subset E$, where $E$ is a finite-dimensional real inner-product space and $0\in\conv U$. Let $A\colon E\to E$ have positive-definite symmetric part, without requiring symmetry, and let $B\geq0$. If each $u_k$ is a $B$-approximate minimizer of $u\mapsto\ip{Az_k}{u}$ over $U$, then $\sup_{k\geq0}\norm{z_k}\leq C(1+\norm{z_0}+B)$, with $C=C(E,U,A)$ independent of the initial state, $B$, and all admissible update choices. We derive two algorithmic consequences. The first is an $O(k^{-1})$ last-iterate norm bound for harmonic vertex-returning Frank–Wolfe for affine strongly monotone variational inequalities on polytopes with relatively interior solutions. It extends the quadratic Frank–Wolfe/herding guarantee of Bach, Lacoste-Julien, and Obozinski \cite[Section~4.2]{BachEtAl2012} to nonsymmetric affine operators. The second consequence concerns oblique relaxation for linear inequalities. Greedy corrections through a fixed matrix with positive-definite symmetric part remain bounded even for inconsistent systems. In particular, for a square matrix $G$ with positive-definite symmetric part, repeatedly increasing the coordinate corresponding to a most-violated inequality of $Gx\geq b$ terminates at an exactly feasible point after finitely many unit corrections. This remains true under bounded additive selection errors, provided the stopping test uses the true inequalities.

When someone asks an AI assistant which doctor to see or which firm to trust with their savings, the answer is a referral. We audit AI provider recommendations in four registry-backed service domains across the 100 largest U.S. metropolitan areas, matching every recommendation against the official registry for its domain (Medicare clinician and facility records, and SEC adviser disclosures), under three conditions: an open-weight model, a proprietary model without web search, and the same proprietary model with search. Without search, both models largely fabricate recommendations in the domains the web covers thinly. Only 4% of the open-weight model's recommended doctors and 11% of the proprietary model's match a clinician in the queried city, and the open-weight matches are name coincidences: its matched clinicians are no likelier to be primary-care doctors than names drawn at random from the registry. With search, 64-71% of recommendations in the same domains match a real provider. Search also changes who is recommended. Without it, recommended advisory firms carry SEC misconduct disclosures at 3.6 times the registry base rate, even after adjusting for firm size; with search, significantly below it. Restaurants, where quality and visibility are separately measurable, show a 3-5x review-count premium but a rating premium of at most a tenth of a star. Finally, search largely removes the metro-size penalty: without it, real recommendations concentrate in the largest metros; with it, match rates are similar across metro-size terciles. Whether an AI referral is trustworthy depends strongly on its retrieval configuration rather than on the underlying model alone, yet an answer produced without retrieval often carries no sign that its recommendations were never verified.

We show that a transverse electric field applied to a zigzag graphene nanoribbon opens a band gap whose energy can be tuned through the terahertz (THz) range. The field breaks the reflection symmetry of the ribbon, allowing optical transitions that are forbidden in its absence. The relative strengths of transitions polarized parallel and transverse to the ribbon depend strongly on excitation frequency and become equal at a particular frequency. At this frequency, each helicity selectively excites carriers in only one of the two field-induced edge-state valleys, with the selected valley reversed for the opposite helicity. Since this excitation occurs away from the valley minima, the photoexcited carriers have finite group velocity, giving rise to a circular photogalvanic response. Van Hove singularities enhance radiative recombination at the band edge, providing a route towards compact, tunable THz emitters.

In this paper, for $N=8,12,16,20,32$, we determine the explicit structure of the graded algebra of modular forms for $\varGamma_0(N)$ of both integral and half-integral weights, with all quadratic Dirichlet characters modulo $N$. We give explicit eta quotient generators and determine the relations among them. Using these structures, for a given weight, a level among those considered above, and a quadratic character, we obtain a basis for the corresponding space of modular forms consisting of eta quotients. Using these eta quotient bases, we obtain necessary and sufficient conditions for the periodicity of the signs of the Fourier coefficients of modular forms in three specific spaces. Finally, we determine all eta quotients in these spaces whose Fourier coefficient signs are periodic.

Adaptive security at the network edge increasingly relies on automated planners, including rule-based controllers, learned policies, and LLM-assisted agents, that translate observations into enforcement actions. Once such a planner can influence live policy state, syntactic validity is not enough. A semantically wrong action, produced from incomplete or manipulated observations, can be faithfully executed by an enforcement substrate that cannot judge mission context. We address this problem by treating the boundary between planner output and kernel enforcement input as the primary security object. We propose a split-control architecture in which an untrusted planner emits typed security intents, a deterministic governor checks each intent against safety, resource, temporal-stability, and proportionality invariants, and only admitted actions are bound to signed receipts and compiled into pre-installed eBPF map updates. The paper formalizes this trust-boundary problem, defines three threat classes, develops the governor admission predicate, and reports an end-to-end prototype. Across rule-based and LLM-assisted planners on a Raspberry Pi 5 testbed connected to the university 5G Test Network, the governor admits, rejects, and bounds intents at microsecond cost without disrupting protected-flow regularity. The contribution is conceptual as much as empirical: adaptive security does not need to trust the author of an action. It needs a mediation boundary that decides whether the action is admissible.

Current methods for the gravitational-wave (GW) echo searches generally require substantial computational time and resources. Recent advances in group equivariant neural posterior estimation (GNPE) have demonstrated the potential to substantially accelerate posterior inference. In this work, we adopt this framework and train a GNPE model to perform echo searches, significantly improving computational efficiency while maintaining comparable inference accuracy. We apply it to the O4a events and find no significant evidence for the GW echo.

Muscle activity is fundamental to human movement, and understanding its patterns is critical for injury prevention and rehabilitation. Conventional muscle activity monitoring relies on specialized sensors such as surface electromyography, which limits its practicality for long-term real-world use. Existing studies suggest that muscle-related information can be inferred from human pose. However, the substantial gap between externally observable pose and internal muscle activation, limits the accuracy and generalization of current approaches. In this study, we propose Pose2Muscle, a pose-driven framework for discrete muscle activity estimation without requiring sEMG signals at inference time. Instead of directly regressing continuous sEMG signals, Pose2Muscle reformulates muscle estimation as a structured prediction problem over discrete muscle activity states, yielding a more stable and interpretable target space. The framework combines multi-scale spatio-temporal attention to capture motion patterns at complementary spatial and temporal scales with a directed acyclic graph-based decoder that maintains multiple candidate muscle-state hypotheses and performs structured trajectory inference over time. To support this task, we construct PoseEMG-43, a synchronized pose-sEMG dataset containing 2,992 movement instances from 43 daily-life actions performed by 14 participants. Experiments show that Pose2Muscle consistently outperforms representative retrieval- and pose-based baselines. It achieves an Adjacent-level Accuracy of 86.36% and a Pearson correlation coefficient of 0.8821 under the Random Split, and 63.97% and 0.6795, respectively, under the Subject-Level Split. These results demonstrate the feasibility of inferring structured muscle-state patterns from human pose and suggest the potential of Pose2Muscle for muscle-aware movement analysis when direct physiological sensing is impractical

We study density functions associated with filtrations of graded ideals in standard graded domains. Extending earlier work, we develop a general theory for arbitrary filtrations and establish their fundamental properties, including existence, continuity, log-concavity, and integral representations. We show that every density function is locally uniformly approximable by piecewise polynomial density functions arising from truncated Noetherian filtrations. We prove differentiability properties of saturation density functions and give a numerical characterization of when the symbolic powers of a graded radical ideal in a polynomial ring coincide with the integral closures of its ordinary powers. Finally, we establish several inequalities for the mixed multiplicities of equigenerated ideals.

Protoplanetary disc dispersal is driven by two processes usually modelled separately: photoevaporative and magnetohydrodynamic (MHD) disc winds. Global simulations indicate that in the inner disc these are not distinct outflows but a single magnetothermal wind. We assemble a closed-form, two-phase model that respects it. A single-field-line wind, whose base is fixed by the irradiated temperature and penetration column, supplies the launch and feeds a secular evolution, with photoevaporation convolved on as a sink. The flux closure $B_z\propto\Sigma^q$ is self-limiting: for $q\le1/2$ depletion alone cannot demagnetise the disc, so dispersal requires independent flux loss, parameterised by the magnetic Reynolds number $\mathcal{R}_m$. Integrating the coupled system yields two regimes. Efficient flux loss ($\mathcal{R}_m\lesssim1$) lets the magnetisation front recede by over an order of magnitude and opens a photoevaporative gap. Flux retention ($\mathcal{R}_m\gg1$) drives the front outward, sustains accretion, and defers dispersal by $\approx2.7$~Myr. Deriving the base from stellar irradiation instead of prescribing it, we find that the cold-launch approximation is valid during the early stages of disc evolution: anchoring the base at plasma equipartition ($\beta_{\rm base} \sim 1$) confines irradiation's influence on the magnetic lever arm to the magnetothermal annulus, decoupling the peak accretion rate from the incident flux. Both regimes clear the disc inside-out, through either a photoevaporatively amplified cavity wall or an expanding magnetothermal front.

Streaming automatic speech recognition (ASR) must be judged jointly on what it transcribes and on how quickly it commits each word. Delayed streams modeling (DSM) has become the dominant paradigm for streaming large audio-language models, exposing a structural delay $\tau$ that bounds the decoder's lookahead. We show that $\tau$ is a poor proxy for user-perceived latency, and that the alignment-based supervision of DSM leaves latency on the table: the same forced-aligned transcript is used at every $\tau$, forcing the model to withhold words it could already commit. We introduce AWED, a word-level emission-delay metric defined relative to the acoustic end of each word, and post-train a DSM recognizer with GRPO under a reward that scores transcription accuracy and measured delay jointly. Trained at a single operating point ($\tau=6$ frames), our model dominates both its supervised fine-tuning initialization and the Voxtral Realtime backbone across all evaluated lookahead budgets: it cuts WER by 30.8% relative at an 80\,ms structural delay, and by 5.7% relative at 480\,ms while lowering median AWED from 1.17\,s to 1.04\,s. Latency-rewarded post-training thus advances the accuracy–latency Pareto frontier of streaming ASR without architectural change.

Phonon angular momentum, characterized by circular or elliptical motion of atoms, plays a decisive role in diverse phenomena, ranging from the phonon Hall effect, phonon magnetic moment, Einstein-de Haas effect, Weyl phonons, and driven chiral phonons. The phonon modes with an antisymmetric Raman tensor, which carry an intrinsic phonon angular momentum in a pseudovector symmetry channel, have been overlooked in the past in the chiral phonon community. Here, we report the detection of this type of phonon mode using polarization-resolved Raman spectroscopy. We observe a phonon at 266cm$^{-1}$ in the antiferromagnetic chiral phase of Cr$_2$O$_3$ in the $A_2$-pseudovector symmetry channel. This $A_2$ mode's frequency and the elliptical motion are captured by first-principles phonon calculations, which treat both lattice and spins on an equal footing as 'slow' degrees of freedom. This mode gets Raman activity due to the presence of Cr vacancies in Cr$_2$O$_3$ that locally break the twofold rotational symmetry perpendicular to the threefold axis. Our results provide a new way to detect the novel symmetry-forbidden $A_2$ phonon modes with an antisymmetric Raman tensor in a pseudovector symmetry channel.

Backman and Liu proved that every integral generalized permutohedron of type $A$, and in particular every matroid base polytope, admits a regular unimodular triangulation. The analogous statement fails in type $B$: the delta-matroid simplex \[\operatorname*{conv}\{\mathbf{0},\ e_1+e_2,\ e_1+e_3,\ e_2+e_3\}\] has normalized volume $2$ and no lattice points other than its vertices, so it has no unimodular triangulation. We show moreover that, up to the natural symmetries of the $0/1$ cube and deletion of constant coordinates, it is the unique non-unimodular delta-matroid polytope that is a simplex. We prove instead that every delta-matroid polytope admits a regular dyadic triangulation, meaning a lattice triangulation whose maximal simplices have normalized volumes that are powers of two. More generally, every integral type $B$ generalized permutohedron admits such a triangulation. The main lattice-theoretic ingredient is that the type $B$ root configuration forms a totally dyadic system, a $2$-local analogue of total unimodularity. As a consequence, these polytopes satisfy a dyadic version of the integer decomposition property. In each dimension the corresponding exponent can be chosen uniformly, even though ordinary integer decomposition can fail for delta-matroid polytopes.

Partizan Edge Geography is a two-player game on a graph where each player has their own token on a vertex and moves their token to a neighbor in a turn removing the edge. Two player alternately move their tokens and the first player who cannot move loses the game. Fraenkel and Simonson (TCS, 1993) showed that the winner determination of this game is PSPACE-complete on directed graphs, given a graph and token positions. This paper resolves its complexity on undirected graphs by showing the PSPACE-completeness on bipartite undirected graphs of maximum degree 3. The same reduction also works for a variant where two tokens cannot be placed on the same vertex.

Remote sensing recognition is often constrained by scarce observations of rare targets and costly annotations, making realistic synthetic augmentation particularly valuable for few-shot and long-tailed scenarios. Object insertion provides an efficient way to increase target diversity while preserving authentic background scenes, but realistic insertion in overhead imagery requires the generated target to adapt coherently to its surrounding environment. To this end, we propose PDA++, a unified environment-aware object insertion framework organized as Plan, Decouple, and Assimilate. Planning determines scene-compatible poses through an affordance field that combines geometric clearance with structure- and scale-aware cues. Decoupling introduces a pose-conditioned background that provides precise spatial guidance together with target-scene context, allowing the reference object to preserve its identity while adapting to the target observation. This construction also naturally provides pixel-level masks for segmentation augmentation. Assimilation further improves local coherence by aligning multi-scale texture distributions through optimal transport. On the optical benchmark, PDA++ achieves a whole-image FID of 6.28 and improves average few-shot recognition mAP50 by 17.69 points, corresponding to a 28.8% relative gain over the real-data baseline. On SAR imagery, it improves ship detection by 4.10 mAP50 points and remains effective under cross-dataset transfer and amorphous-target insertion. Code is available at https://github.com/lisheyu972/PDA_PLUS.

Recent advances in LLMs and VLMs have enabled safety systems to reason beyond simple risk patterns toward more contextual and semantic safety concerns. However, as risk patterns continue to evolve and safety rules become more complex, existing training-based end-to-end safeguards face persistent challenges in adaptability and explainable reasoning over complex safety rules. To address these challenges, we propose GuardEn (Guarding by Safety Rule Entailment), an executable safeguard framework that decomposes safety policies into atomic propositions through Safety-Rule Compilation, modeling their composition as executable code. At test time, Scene-Grounded Execution instantiates these atomic propositions with contextual visual information derived from scene graphs, enabling rule-grounded and interpretable safety reasoning. Experiments on SafetyVisionBench demonstrate the effectiveness of programmable safeguard for complex visual safety assessment, achieving an average improvement of 9.8 F1 points over the strongest baseline.

A framework for quantum electrodynamics of equilibrium systems is proposed that takes the Maxwell equations as foundational and links them to the structure of density-functional theory for the quantum system. By switching from purely internal observables like the one-particle density to combined internal and external quantities, the external sources and their energy are taken into account. A fully regularized functional theory emerges that avoids the usual representability problems. With the Kohn-Sham construction that links to an auxiliary uncoupled system, the ultra-violet cutoff can be removed and one arrives at a fully renormalized and non-perturbative light-matter description. Regularized forms of density-functional theory with and without magnetic fields appear as boundary cases, while the emergent physical picture reproduces the macroscopic Maxwell equations.

Unmanned aerial vehicles (UAVs) extend embodied intelligence into continuous three-dimensional space, where perception, reasoning, physical embodiment, and action are tightly coupled through flight and environmental interaction. Recent advances in foundation models, world models, and AI agents are shifting UAV autonomy from task-specific perception and control toward systems that can interpret human intent, understand open environments, reason about physical consequences, and organize complex behaviors under embodiment and flight-dynamic constraints. We characterize this emerging paradigm as UAV embodied intelligence (UAV EI) and distinguish it from its system realization, the embodied-intelligent UAV (EI UAV). To provide a unified view of the field, we introduce a 5+5 framework that describes UAV EI through five capability dimensions and EI UAVs through five architectural layers spanning physical embodiment, general cognition, embodied skills, external interaction, and system harnessing. Based on this framework, we systematically review recent progress in embodied morphology, embodied perception, world models, embodied planning, vision-language navigation, embodied manipulation, and embodied collaboration. We further identify long-horizon autonomy, predictive physical reasoning, test-time skill acquisition, and autonomous capability evolution as key challenges toward more general aerial embodied intelligence. Finally, we argue that harnessing physical-digital AI agents, through persistent coupling of digital intelligence with physical sensing, dynamics, action, and feedback, provides a system-level pathway toward adaptive and continuously evolving UAV autonomy. Project resources are available at our project website and GitHub repository.

This paper focuses on a class of nonconvex nonsmooth multicomposite optimization problems for training RNNs (Recurrent Neural Networks). We first establish easily verifiable conditions under which an approximate first-order d-stationary point of the problem is guaranteed to be an approximate second-order d-stationary point. Subsequently, we propose a sequential regularized piecewise affine algorithm, SRPA, that minimizes a regularized piecewise affine function at each iteration, where the idea of active-set strategy is incorporated to reduce the computational cost per iteration. Leveraging the aforementioned conditions for second-order d-stationarity, we establish the global convergence of SRPA to second-order d-stationary points and the complexity bound of $ \mathcal{O}(\epsilon^{-2}) $ for obtaining an $ \epsilon $-approximate second-order d-stationary point. Finally, numerical experiments for training RNNs on synthetic and real-world datasets demonstrate the promising performance of SRPA compared with state-of-the-art algorithms.

Detecting failures during execution is essential for reliable robotic manipulation. Vision-language models (VLMs) can assess task outcomes semantically but add runtime computation, whereas out-of-distribution (OOD) detectors may respond to harmless scene variations rather than failure-relevant deviations. We introduce RAFAIL, a framework for detecting execution failures during robotic manipulation. RAFAIL identifies failures by detecting anomalies in task-relevant relationships between entities, such as a gripper and an object or an object and its target. By focusing OOD detection on relevant parts of the observation, RAFAIL reduces sensitivity to task-irrelevant scene variation. Offline, a VLM annotates successful demonstrations with task progress and relationship importance, which are used to learn point-cloud-based relationship representations without relying on policy-internal features. At runtime, relationship-specific OOD detectors evaluate these representations while relationship importance and task progress are predicted without VLM inference. RAFAIL requires no failure data and achieves 73.4% balanced accuracy across three real-world robotic manipulation tasks, outperforming the strongest evaluated OOD- and uncertainty-based baselines.

Recent Omni-Modal Generative Models (Omni-Models) have advanced content generation toward unified modeling of text, images, video, and audio. MiniMax-H3 exemplifies this transition by combining multimodal context understanding with joint audio-visual generation in a shared latent framework. Its unified architecture raises a fundamental question: Can multimodal alignment improve the model's world reasoning, and what new evaluation paradigms do omni-modal inputs enable? To investigate this question, this work introduces a comprehensive evaluation framework organized around four complementary dimensions of physical world reasoning. Unlike existing evaluation frameworks for video generation and world models, which are often constrained by limited input modalities and evaluation settings where prompts closely match the target video content, our evaluation is specifically designed to exploit the multimodal inputs of Omni-Model. We construct a diverse set of novel tasks that require models to integrate complementary information across modalities. Specifically, we consider four scenarios, including implicit prompts paired with multiple frames, audio-image, prefix-videos, and audio-video inputs. Every single modality provides only partial evidence about the underlying event, requiring the model to jointly reason over the complementary semantic cues to infer latent event states and future dynamics. Across 517 evaluation instances, MiniMax-H3 achieves an overall success rate of 41.97%. Video-based Decision Reasoning yields the highest success rate at 56.00%, while Audio-based Disambiguation Reasoning is the weakest, reaching only 27.40%. These results indicate that effective multimodal integration remains key to fully exploiting the benefits of diverse input modalities. The project is available at https://github.com/gulucaptain/MiniMax-H3-Reason.

The practical use of statistical depth for fuzzy data requires regularity, inferential stability, and computational feasibility. This paper studies these aspects for the first time for the fuzzy depth, in particular for the fuzzy Tukey depth. In particular, we investigate the continuity of this depth with respect to each of its arguments. As a function of the elements of the underlying space, we prove its upper semicontinuity under the main metrics used in fuzzy spaces. As a function of the distribution with respect to which the depth is computed, we establish the almost sure uniform consistency of its empirical version for continuous fuzzy random variables. These results ensure closed depth regions and support the use of sample depth values to approximate their population counterparts. We also derive consistency of empirical deepest-point estimators and study a finite-grid approximation of the depth, supported by theoretical and simulation results.

Offline on-policy distillation gains efficiency by collecting student trajectories and teacher supervision once and reusing them throughout optimization. The same reuse makes imperfect supervision persistent. Since even strong teachers can fail, we ask what remains learnable from imperfect teacher supervision? Teacher failure is only a coarse problem-level signal and does not imply that all supervision along the associated student trajectory is unhelpful. A natural alternative is to estimate teacher recoverability along the trajectory, but repeated continuations largely erase the efficiency advantage of offline distillation. We instead use teacher-successful problems to define a cheap reference for what the student can learn. We train on teacher-successful problems and measure how the likelihood of each observed token in trajectories from teacher-failed problems changes. We use these signed likelihood changes as an operational learnability signal: larger increases indicate behavior more strongly promoted by successful-only learning. We aggregate this signal into trajectory-level weights for the original distillation loss. Unlike continuation-based estimates, our learnability requires no additional generation and can be computed once from stored trajectories and model checkpoints. Across mathematical reasoning and code generation, our method improves an offline OPD baseline by up to 2.7 percentage points and matches or outperforms online OPD variants on multiple benchmarks. Despite the additional successful-only distillation stage, it uses 2 GPUs and about 22 GPU hours, compared with 3 GPUs and 36–48 GPU hours for representative online OPD methods.

Large Language Models (LLMs) frequently exhibit hallucinations, presenting a major barrier to reliability in complex reasoning tasks. While traditional detection methods rely on output-based confidence metrics, these logits are often miscalibrated by modern alignment techniques. In this paper, we investigate the temporal volatility of internal attention mechanisms as an alternative diagnostic signal for hallucination that does not depend on output calibration. By introducing an unsupervised metric for attention dispersion, we show that epistemic uncertainty leaves a measurable trace within intermediate layers, where spikes in attention entropy are associated with reasoning breakdowns. We evaluate our approach on mathematical reasoning benchmarks (GSM8K and MATH-500) using the Qwen2.5 model family (1.5B and 3B parameters), finding statistically significant AUC improvements of up to +0.076 over output-based baselines across all tested conditions. These findings suggest that attention dispersion is a promising complement to traditional hallucination detection methods, requiring further investigation across broader model families and task domains.

Assessing the cosmological typicality of the Local Group is crucial for establishing whether its dwarf galaxy population is representative. To this end, the Delegate Survey targets dwarf galaxies in Local Group analogues for systematic comparison. As a pilot study for the Delegate Survey, we obtained integral-field spectra of three dwarf galaxies, GAMA 79098, GAMA 569709, and HIPASS J1159-19 S2, in two nearby galaxy groups using the KOALA instrument on the AAT. The galaxies were mapped at a spatial sampling of 1.25 arcsec per spaxel, providing two-dimensional spectral coverage of both the stellar and ionised gas components. Using continuum and emission-line maps, we investigate the ionisation structure, star-forming properties, metallicity, star formation histories, and velocity fields of each system. GAMA 79098 exhibits properties closely analogous to the SMC, with ongoing star formation superimposed on an old stellar population and moderate rotational support. GAMA 569709 resembles transition-type Local Group (LG) dwarfs such as NGC 147 and NGC 185, showing predominantly an old stellar population. HIPASS J1159-19 S2 is dominated by young stars and intense star-forming complexes, consistent with a dynamically young, gas-rich galaxy group environment. The two GAMA dwarfs follow the LG stellar mass-metallicity relation within uncertainties, while HIPASS J1159-19 S2 appears approximately 0.5 dex more metal-rich for its stellar mass. Overall, the structural, chemical, and kinematic properties of these dwarfs overlap substantially with analogous LG dwarfs.

We use $m$-dimensional partial spreads of $\fp^{2n}$, where $p$ is an odd prime, $n$ is a positive integer and $m$ divides $n$, to construct two classes of bent functions from $\fp^{2n}$ to $\fp$. Our construction generalizes the classes of $p$-ary $\cPS^{-}$ and $\cPS^{+}$ bent functions proposed by P. Lison\v ek and H. Y. Lu (Des. Codes Cryptogr. 73 (2014), 209–216).

Large language models (LLMs) suffer from a long-tail deficit: culturally specific facts, particularly those concerning underrepresented regions such as Latin America, appear too rarely in pretraining corpora to be reliably memorized. Retrieval-Augmented Generation (RAG) addresses this by grounding generation in external text, but structured alternatives such as Knowledge Graphs (KGs) offer tighter control over what enters the context, along with potential gains in explainability and updatability. We benchmark Graph-RAG against standard RAG on LatamQA, a culturally grounded multiple-choice dataset spanning eight thematic categories. The graphs are built end-to-end from Wikipedia articles with KGGen, a recent open-domain extractor, without manual curation in our main setting. G-Retriever is competitive with RAG and reduces the error of the base LLM by 72% with a standard KG and 78% with a benchmark-aware variant, the gap to RAG narrowing further as the graph is oriented toward task-relevant content. The trained projection transfers zero-shot to Portuguese without target-language fine-tuning, indicating multilingual reach.

Modern numerical astrophysics applications present a common demand for efficient spatial querying techniques, such as neighbor searches or density estimations over billions of resolution elements. To address the growing need for fast and versatile tools tailored to these tasks, we present cosmokdtree, a fast and flexible multi-purpose k-d tree implementation in Fortran, parallelized with OpenMP directives. Our library supports arbitrary dimensionality and spatial distributions, and is designed for efficient tree construction and fast query performance for moderate-scale applications (performance tested up to 10^9 points). Python bindings coupled to the module are also provided. All these ingredients yield a k-d tree package suitable for analysis purposes. We benchmark cosmokdtree across a variety of scenarios, including different point distributions, dimensionality, and parallel scaling. We also compare its performance against the widely used scipy implementation and other efficient alternatives such as coretran's and a GPU-based version, showing that cosmokdtree consistently achieves lower construction and query times than CPU alternatives while keeping a reasonable memory usage. Our tree-building phase implementation, executed on a mid-range workstation-class CPU, approaches the performance of GPU-based implementations when run on high-end consumer graphic cards, although data-center GPUs, out of the scope for our comparison, could still deliver substantially higher performance and a broader conclusion cannot be extracted. We further present some applications tackling common problems in astrophysics, namely, the friends-of-friends clustering algorithm and the particle-to-mesh assignment process. The code is publicly released and intended to serve as a flexible multi-purpose tool for computational applications in a wide range of scenarios, particularly in astrophysics.

Non-intrusive load monitoring (NILM) estimates appliance power sequences from aggregate power, but models trained on source households commonly lose accuracy in unseen households. Aggregate power also contains loads from other appliances and measurement error, so predictions may depend on the residual background that co-occurs with source-household targets. Time-aligned submetered measurements and the additive decomposition of aggregate power expose a relation unused by window-wise supervision: an aggregate window can be recomposed by replacing only its residual background while preserving all modeled target-appliance power sequences pointwise. We combine label-preserving aggregate recomposition with prediction consistency. Both windows receive complete power and operating-state supervision. For each appliance, disagreement between the two power predictions is penalized only when both satisfy a fixed reliability criterion and only to the extent that it exceeds a fixed margin. The proposed method is implemented using a multi-appliance architecture with two-stage shared-to-specific mixture-of-experts routing. On REDD, UK-DALE, and REFIT, the proposed method lowers appliance-averaged mean absolute error relative to single-window training from 14.75 to 13.14 W, from 8.88 to 8.51 W, and from 15.83 to 14.55 W. Label-preserving aggregate recomposition and prediction consistency are used only during training, and add no inference-time module or parameter.