Astrophysics of Galaxies
The gradual decline of Ly$\alpha$ visibility in the CANDELS fields: evidence for the combined effects of galaxy evolution and reionization
L. Napolitano, L. Pentericci, A. Ferrara et al. (+35 more)
We investigate the
evolution of Ly$\alpha$ visibility and the physical properties of Ly$\alpha$ emitters (LAEs) across the five CANDELS fields using publicly available JWST/NIRSpec PRISM spectroscopy. Our catalog comprises 3446 spectroscopically confirmed sources at 4 $\leq $ z < 14.2, including 3361
star-forming
galaxies (SFGs), of which 539 are robust (S/N > 3) LAEs. We measure the fraction of LAEs with EW$_0$ > 25A (X$_{\mathrm{Ly\alpha}}$) and trace its
redshift evolution in two UV luminosity bins, namely -20.25 < M$_{\mathrm{UV}}$ < -18.75 and -21.75 < M$_{\mathrm{UV}}$ < -20.25. Within the fainter-UV range, X$_{\mathrm{Ly\alpha}}$ increases from z = 5 to z = 6 at 3$\sigma$ significance and subsequently declines toward higher
redshifts with a significant monotonic trend at z > 6. We also investigate the physical properties of both LAEs and the full SFG population. We find that the
stellar mass, UV slope $\beta$,
stellar reddening, SFR,
metallicity, sSFR, and burstiness of LAEs remain approximately stable with
redshift. The only exception is the mass-weighted age which decreases with increased
redshift. Conversely, the properties of the full SFG population evolve significantly, progressively approaching the region of
galaxy-property space occupied by LAEs as
redshift increases. This suggests that
galaxy evolution may enhance the intrinsic production and escape of Ly$\alpha$
photons toward earlier epochs. We argue that this effect should be accounted when inferring the
evolution of the neutral hydrogen content of the IGM from the observed visibility of Ly$\alpha$ emission. To this end, we employ a physically motivated framework based on the Attenuation-Free Model, jointly accounting for
galaxy evolution and IGM attenuation. Our observations favor reionization histories that proceed gradually over scenarios characterized by a rapid increase in the
cosmic neutral hydrogen fraction.
Read the full abstract ▾
All authors L. Napolitano, L. Pentericci, A. Ferrara, M. Llerena, M. Dickinson, A. Calabrò, S. L. Finkelstein, R. Begley, P. Arrabal Haro, A. Arroyo-Polonio, B. E. Backhaus, D. Bevacqua, A. Bhagwat, M. Bischetti, M. Castellano, S.-J. Chang, E. R. Cueto, V. D'Odorico, C. T. Donnan, M. Galbiati, G. Gandolfi, M. Giavalisco, M. Hirschmann, J. Kartaltepe, A. M. Koekemoer, K. K. Knudsen, R. A. Lucas, S. Mascia, L. Paquereau, B. Pérez-Díaz, C. Piscitelli, R. Rana, P. Santini, A. J. Taylor, E. Taylor, R. Tripodi, S. M. Wilkins, L. Y. A. Yung
Computer Vision
A Lagrangian View of Flow Matching
Peyman Milanfar
Modern explicit-time generative models, such as Flow Matching [Lipman et al., 2023] and Rectified Flow [Liu et al., 2023], are typically derived top-down via Optimal Transport and the continuity equation. This standard Eulerian approach focuses on the macroscopic transport of probability mass. In this paper, we present an alternative, bottom-up mechanical derivation grounded in a Lagrangian (particle-centric) perspective. By analyzing the local Taylor expansion of a continuous denoiser, we motivate a strict invariance condition required for optimal, singlestep generation: the conservation of target identity. Enforcing this condition yields a governing quasi-linear advection Partial Differential Equation (PDE). We demonstrate that solving this PDE via the Method of Characteristics analytically yields the straight-line trajectories of Flow Matching. This
geometric perspective isolates the Jacobian of the denoiser as the primary source of trajectory
curvature, providing a direct mathematical explanation for why straight-line flows enable massive step sizes, and why empirical models require
distillation to flatten intersecting characteristics.
Read the full abstract ▾
Social and Information Networks
Social bots weaken activist cohesion
Linda Li, Orsolya Vasarhelyi and Balazs Vedres
Social bots now make up a substantial share of online political communication, where they are studied mainly as producers of misinformation and amplified content. Far less is known about whether their presence reshapes the human relationships that hold movements together. We ask whether exposure to bots during a protest peak is followed by the erosion of cohesion in human networks. Tracking retweet networks of core participants in the 2020 Black Lives Matter (BLM) protests before, during, and after the peak, we measure change in cohesion at two scales: triadic closure in individual ego networks and edge density within detected communities. Greater bot exposure during the peak predicts steeper subsequent declines in human cohesion at both scales, and the loss concentrates among supporters of the movement. Bots may weaken activism less by changing what people believe than by dissolving the ties through which collective action is sustained.
Read the full abstract ▾
Machine Learning
WHALE: A Simple Recipe for Joint Harness-Weight Optimization
Haechan Kim, Yoonho Lee, Gisang Lee et al. (+2 more)
Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness updates can change which model capabilities are exposed. Existing joint-adaptation methods optimize weights and textual
prompts but leave the broader harness fixed. We propose Weight-Harness Alternating LEarning (WHALE), a simple recipe that alternates two phases: updating the model under the current harness, then searching for a better harness under the updated model. We instantiate these two phases with online rejection-sampling fine-tuning and Meta-Harness, respectively. When to switch is a key design choice: to separate real improvements from noise without over-optimizing against a changing counterpart, WHALE uses either fixed phase durations or an adaptive patience rule over training signals. Using Qwen3.5-2B/4B
agents across three domains (search question answering, mathematical
reasoning, and chess puzzles), WHALE outperforms weight-only, harness-only, and Fast-Slow Training by 4.15-24.38 percentage points in best mean@8 accuracy. Either component can be the bottleneck: harness search matches peak weight-only accuracy with far fewer rollouts in SearchQA, but improves math accuracy only after a weight update. Small interleaved updates also outperform stagewise weight-then-harness optimization in accuracy and rollout cost. The code is available at https://github.com/krafton-ai/WHALE.
Read the full abstract ▾
All authors Haechan Kim, Yoonho Lee, Gisang Lee, Chelsea Finn, Kangwook Lee
Human-Computer Interaction
Cyber-Physical Digital Factory Architecture as the Enabler of Disembodied Work
Tero Kaarlela, Ivan Ruchkin, Jose Outeiro et al. (+1 more)
Digital Twins (DTs), Artificial Intelligence (AI), and Industrial Internet of Things (IIoT) technologies have significantly advanced manufacturing digitalization. However, these technologies are typically applied to individual manufacturing processes rather than integrated into a unified cyber-physical manufacturing environment. This paper proposes a cyber-physical digital factory architecture that enables disembodied work, where manufacturing systems can be supervised and operated remotely through eXtended Reality (XR) user interfaces in collaboration between AI-based control and human operators. The architecture integrates synchronized DTs, hierarchical cloud-edge AI, IIoT, and XR teleoperation interfaces into a cyber-physical manufacturing environment. The proposed approach is validated through representative manufacturing operations, including CNC machining, robotic-assisted abrasive finishing, and robotized disassembly. The results demonstrate the feasibility of the proposed architecture for disembodied manufacturing work and provide a reusable cyber-physical framework for future human-AI-controlled digital factories.
Read the full abstract ▾
All authors Tero Kaarlela, Ivan Ruchkin, Jose Outeiro, Souradeep Dutta
Artificial Intelligence
ReDeck: Step-Level Render-Grounded Refinement for Document-to-Slide Generation
Muzhao Tian, Zezi Zeng, Yifan Yang et al. (+14 more)
Document-to-slide generation is challenging because slides are dense editable artifacts that require both faithful content selection and precise spatial layout. Recent slide
agents adopt iterative reflection, but typically follow a monolithic "one version, one feedback" loop: a slide or deck is rewritten, rendered afterward, and critiqued only at the turn boundary. This delayed feedback makes local failures such as overflow, overlap, clipping, and off-canvas placement difficult to attribute and repair. We propose ReDeck, a step-level render-grounded refinement framework that decomposes slide revision into atomic edit actions and returns renderer-derived observations after each step, turning refinement into "one edit, one observation." To balance local repair with global quality, ReDeck uses multi-granular feedback: step-level render feedback for spatial errors, a turn-level adaptive critic for semantic and design guidance, and a submission-level gate for hard layout validation. We further introduce DeckQuiz, a
benchmark that decouples content fidelity, spatial correctness, and design quality. Across GPT-5.4, Claude-4.6, and Gemini-3.1, ReDeck consistently outperforms existing slide-generation
agents, and ablations confirm that feedback timing and granularity are critical for reliable slide refinement.
Read the full abstract ▾
All authors Muzhao Tian, Zezi Zeng, Yifan Yang, Xin Gao, Yan Li, Zisu Huang, Xiaohua Wang, Changze Lv, Mingxi Cheng, Bei Liu, Kai Qiu, Qi Dai, Dong Chen, Yue Dong, Xiaoqing Zheng, Ji Li, Chong Luo
Machine Learning (Statistics)
Provably Efficient Federated Reinforcement Learning with Linear Function Approximation and Logarithmic Communication Cost
Zihang Liang, Haochen Zhang and Lingzhou Xue
We study
federated online reinforcement learning with linear function approximation. While recent multi-
agent reinforcement learning
algorithms achieve strong regret guarantees, they typically require sharing raw trajectories. This reliance incurs a communication cost that scales linearly with the number of episodes and violates the
privacy constraints of
federated settings. To address these limitations, we propose Fed-LSVI, the first provably efficient
federated algorithm for online reinforcement learning with linear function approximation in episodic Markov decision processes. By integrating a determinant-based event-triggered synchronization with a stepwise backward update mechanism, Fed-LSVI enables
agents to collaboratively learn an optimal policy by exchanging only compressed sufficient statistics. We prove that Fed-LSVI achieves a regret bound of $\widetilde{\mathcal O}(\sqrt{Md^3H^4T})$, where $d$ is the feature dimension, $H$ is the horizon length, $M$ is the number of
agents, and $T$ is the number of episodes per
agent, matching the best-known regret for multi-
agent online reinforcement learning with linear function approximation. Moreover, by following the stringent communication and
privacy constraints of the
federated setting, Fed-LSVI reduces the communication cost to only logarithmic dependence on $T$, representing a significant improvement over prior methods.
Read the full abstract ▾
Artificial Intelligence
LLM-Driven Autonomous Vehicles Can Inherit Human Driver Biases in Pedestrian Yielding: Results and Implications From a New Benchmark
Irem Yoldas, Jie Zhang, Odinaldo Rodrigues et al. (+1 more)
Public trust in Autonomous Vehicles (AVs) may depend not only on technical success but also on the
fairness of their decision making. While a recent trend in AV research involves using general purpose "common sense" models to guide AV decision making, the degree to which these inherit human biases in driving is still understudied. Given that psychology studies have shown human driver biases exist, such as lower pedestrian-yielding rates to Black pedestrians in the US, we argue that analyses of model bias should also be part of AV evaluation. Concretely, in this paper we propose two new bias testing methodologies for Large
Language Models (LLMs) and Visual-
Language Models (VLMs), "All Else Being Equal" tests and "Self-Consistency" tests, and a new
benchmark (FairYield) to assess bias in pedestrian-yielding decisions. Our findings show that the yielding decisions of both LLMs and VLMs can be influenced by models' predictions of pedestrian gender, ethnicity, religion, disability, age, skin tone, and socio-economic status. While the type and degree of bias is different from model to model, we highlight common patterns, and raise questions about the "common sense" model paradigm, particularly the need to either revise the paradigm or address issues of downstream bias.
Read the full abstract ▾
All authors Irem Yoldas, Jie Zhang, Odinaldo Rodrigues, Martim Brandão
Computation and Language
Assessing Suicide Risk in Arabic Crisis Helpline Calls: A Comparison of Arabic and English Large Language Models
Linhai Ma, Rita El Hachem, Mahatab El Hajj et al. (+2 more)
Crisis helplines assess suicide risk through structured interviews, a process that is slow and dependent on operator training and workload. Natural
language processing could support risk assessment and call prioritization, but almost no work addresses Arabic-
language helpline calls or operates within the
privacy constraints of real helpline data. We analysed de-identified transcripts from Lebanon's National Lifeline for Emotional Support and Suicide Prevention. Audio never left the helpline: calls were transcribed on site with a
speech recognition model for Levantine Arabic, and an Arabic named-entity recognition model removed identifying information locally. Only the de-identified transcripts were shared with the research team. Operators recorded the five suicidal ideation items of the Columbia Suicide Severity Rating Scale, which we combined into two binary outcomes: at-risk and high-risk. We also machine-translated the transcripts into English, giving a paired Arabic/English comparison. On each corpus, we fine-tuned five instruction-tuned large
language models alongside six
transformer encoder baselines (four Arabic, two English) and evaluated all models on a held-out test set. We included 383 calls: 373 for the at-risk task (52.3% positive) and 297 for the high-risk task (30.0% positive). The best Arabic model reached a macro-F1 of 81.19 and a ROC-AUC of 90.61 on high-risk; the best English model reached 85.00 and 92.59, identifying 88.9% of high-risk calls. In both
languages, high-risk calls separated more cleanly than at-risk calls, and translation to English did not reduce the best observed performance. Suicide risk can be classified from de-identified Arabic transcripts without sending audio outside the helpline. The high-risk results support further testing as an operator-facing tool; lower-severity ideation proved the harder case.
Read the full abstract ▾
All authors Linhai Ma, Rita El Hachem, Mahatab El Hajj, Lilian Ghandour, Samah Fodeh
High Energy Astrophysical Phenomena
GRMHD Simulations of Accreting Proto-Magnetars II. Implications for r-process Nucleosynthesis
Tejas Prasanna, Kyle Parfrey, Brian D. Metzger et al. (+3 more)
Newly formed, rapidly spinning, strongly magnetized
neutron stars ("millisecond proto-magnetars") can arise in collapsars,
neutron star mergers, or white-
dwarf accretion-induced collapse, and are often surrounded by compact
accretion disks. At
accretion rates of ~0.1 Msun/s, these disks can become
neutron rich and power
outflows capable of rapid
neutron-capture (r-process) nucleosynthesis. In Paper I, we presented axisymmetric GRMHD simulations of
accretion onto such proto-magnetars and showed how the disk-magnetosphere interaction regulates jet power, variability, and
neutron star torques. Here we use the same simulations to study how this interaction regulates the mass, composition, and velocity of the
baryon-rich ejecta, comparing proto-magnetar models to otherwise similar black hole
accretion. A magnetized
neutron star qualitatively changes both the amount and composition of the ejecta. Stronger
neutron star magnetic fields suppress
accretion and redirect more inflowing material into unbound
outflows, even when the magnetosphere remains strongly compressed by the disk. Once the field produces
magnetic channeling or centrifugal acceleration, mass loss is enhanced further. Reaction-network calculations show that these magnetically driven
outflows can synthesize the full range of r-process
nuclei, including the heaviest elements. Moderate
neutrino irradiation substantially reduces the third-peak yield, but magnetically accelerated
neutron star outflows retain a heavy component more effectively than black hole disk winds; sufficiently strong early-time irradiation suppresses it altogether. Accreting proto-magnetars may therefore be important heavy r-process sources once their
neutrino emission has sufficiently declined.
Read the full abstract ▾
All authors Tejas Prasanna, Kyle Parfrey, Brian D. Metzger, Ore Gottlieb, Andrei Beloborodov, Koushik Chatterjee
Machine Learning
Elite-Weighted Supervised Fine-tuning for Goal-Directed Molecular Optimization
Shiyun Wa, Yifei Wang, Anna G. Green et al. (+2 more)
Goal-directed optimization is essential for steering molecular generators to propose candidates with desired properties. However, it is often implemented with policy-gradient reinforcement learning, which requires a generation-trajectory log-probability whose form depends on the model architecture and generation procedure. This makes an optimizer difficult to reuse across architectures and conditional generative designs. Supervised fine-tuning needs none of that machinery, but its update is driven by a fixed
dataset, so the reward never enters the update. We introduce Elite-Weighted Supervised Fine-tuning (EW-SFT), which uses reward to guide elite selection of high-scoring molecules, and updates the model by its own pretraining loss on that set. Ablations show that reward information is passed primarily through elite selection, rather than through continuous weighting within the selected set. Because the update consumes only scored molecules and the model's native loss, the same rule applies across autoregressive, masked-
diffusion, and discrete-flow generators, and across de novo, motif-extension, and linker-design tasks. Under a fixed budget of 3D shape
alignment oracle calls on two kinase reference compounds, EW-SFT consistently outperforms the corresponding native optimizers. It further improves goal-directed optimization under a 2D similarity oracle on four held-out references and achieves comparable performance on a sample-efficiency
benchmark without a trajectory-level RL formulation. These results demonstrate that EW-SFT is a unified and effective optimizer across molecular generators, design constraints, references, and oracles.
Read the full abstract ▾
All authors Shiyun Wa, Yifei Wang, Anna G. Green, Simone Sciabola, Ye Wang
Computer Vision
ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training
Xionghao Wu, Yijun Yang, Shiyang Zhou et al. (+17 more)
Robotic manipulation faces a fundamental scaling challenge: robust generalization demands broad physical experience, yet action-labeled
robot trajectories are expensive to collect and inherently limited in diversity.
Egocentric videos offer a far more scalable source of
embodied experience, capturing object interactions, contact dynamics, tool use, and long-horizon behaviors across diverse environments. The central challenge is how to convert this abundant but action-free experience into effective
robot control. We introduce ZimaBlue, a scalable framework for learning generalizable World Action Models (WAMs) from large-scale
video. ZimaBlue follows a three-stage training curriculum: it first performs causal
embodied video pre-training on large-scale human and
robot egocentric videos, then grounds the learned visual dynamics in heterogeneous
robot trajectories through
video-action mid-training with a unified action representation, and finally specializes the model to a target
robot for deployment. To make generative WAMs practical for real-time control, ZimaBluefurther adopts an asynchronous Slow-Fast dual-system architecture, where a high-capacity Slow world model provides generalizable spatiotemporal representations and a lightweight Fast branch enables 30 Hz action prediction on NVIDIA RTX 4090. On real-
robot zero-shot evaluations, scaling from target-
robot data alone to over 120,000 hours of
embodied video improves success from 36.1% to 77.8%. ZimaBlue further delivers strong performance across multiple
benchmarks, with particularly pronounced gains on unseen tasks.
Read the full abstract ▾
All authors Xionghao Wu, Yijun Yang, Shiyang Zhou, Haoze Sun, Jianhui Liu, Songsong Yu, Jiyao Zhang, Wenbo Li, Bo Wang, Guoqing Ma, Lin Song, Renjie Liao, Shenghe Zheng, Wei Tang, Xiaojuan Qi, Yanwei Li, Yuan Zhang, Zhuotao Tian, Haoyang Huang, Nan Duan
Computational Engineering
Ordinal Gates, Cardinal Bets: Matching LLM Confidence to the Financial Decision Operator
Rayansh Singh and Sara Rezaeimanesh
LLM confidence scores are not independently deployable objects: their decision value depends on the downstream operator and exposure controller that consume them. Monotone recalibration cannot change a coverage-matched rank-based gate, whereas position sizing consumes score magnitude, so changing a confidence map can invalidate a scale fitted to the previous score distribution. We test this on FactSet news for Nasdaq-100 equities, fitting maps and scales on 2021 and evaluating nine open-weight LLMs out-of-sample on 2022–2023. Cross-applying raw and correctness maps with independently fitted scales shows that the two components are not portable alone: scale transfer reduces certainty-equivalent return (CER) in $8/9$ models and produces large risk-target errors. Matching each map with its fitted scale improves ensemble CER by $9.2$ percentage points per year under frozen-scale control ($p<0.001$), and the effect remains significant when the single largest-contributing model is excluded ($+5.5$pp/yr), so it is not driven by one case. Under an identical adaptive-
volatility controller, however, the incremental effect falls to $+1.6$pp/yr, with a significant controller interaction. Annual walk-forward effects are smaller, although map–scale interaction remains positive in every fold. Confidence transformations should therefore be evaluated jointly with the downstream controllers that consume them.
Read the full abstract ▾
Differential Geometry
Macroscopic scalar curvature bounds on surfaces
Otis Chodosh and Noa Vikman
Algebraic Geometry
Nonexistence of degree two rational multisections of conic bundles over the plane
Jeffrey Diller, Lena Ji and Eric Riedl
We prove that a standard conic bundle $X \to \mathbb P^2_{\mathbb C}$ whose discriminant is very general of degree at least 18 admits no rational multisections of degree two. This is the first step towards proving a
conjecture of Iskovskikh that there are conic bundle threefolds that are not unirational, since to prove that $X$ is not unirational, it suffices to show that there are no rational multisections of any degree. Proving Iskovskikh's
conjecture would provide the first example of a rationally connected variety that is not unirational.
Read the full abstract ▾
Computation and Language
Parallel Universes for Temporal Evaluation and Knowledge Updating in LLMs
Jonathan Zheng, Zirui Shao, Alan Ritter et al. (+1 more)
Large
language models (LLMs) rely on static pretraining corpora, resulting in outdated knowledge over time. Existing approaches for evaluating knowledge edits either suffer from rapid contamination or rely on counterfactual edits that conflict with rigid existing knowledge. In this work, we propose a simulation-driven approach for evaluating and training LLMs on knowledge insertions. We introduce PARALLELEVENTS, a
benchmark that simulates coherent trajectories of fictional yet realistic future events, enabling controlled evaluation without data contamination or conflicts with existing knowledge. We further propose SYNAPSE, a training framework that leverages model-generated data to integrate new knowledge into model parameters via instruction tuning. This synthetic pipeline enables scalable knowledge integration without costly human-curated data. SYNAPSE improves QA accuracy by 14.23 points over prior methods, indicating that simulation-based synthetic training yields robust, coherent knowledge insertion.
Read the full abstract ▾
All authors Jonathan Zheng, Zirui Shao, Alan Ritter, Wei Xu
Differential Geometry
Rectifiability of the Singular Strata for Harmonic Maps to DM-Complexes
Ivy Stoner and Yitong Sun
We prove a rectifiability
theorem for harmonic maps into DM-complexes. More specifically, if $u$ is a harmonic map from a $n$-dimensional Riemannian domain to an $N$-dimensional NPC DM-complex $Y$, the $k$-th singular stratum of the singular set is countably $k$-rectifiable for all $k \in \{0,...,n-2\}.$ Our proof follows the framework in \cite{dm}. This extends the rectifiability result for the singular set of harmonic maps into a $F$-connected complex in \cite{dees} and generalizes the rectifiability
theorem with respect the singular strata of harmonic maps into a $F$-connected complex in \cite{bd}.
Read the full abstract ▾
Astrophysics of Galaxies
Local Heavily Obscured Active Galaxies Observed With Chandra (I): Spectral Properties of the Nuclear and Extended X-ray Emission
R. Middei, G. Fabbiano, A. Trindade Falcão et al. (+3 more)
The Chandra X-ray Observatory has revealed that kiloparsec-scale extended X-ray emission is commonly associated with heavily absorbed ($\log (N_{H}$/cm$^{-2})\,>\, 23$) Active
Galactic Nuclei (AGN) in the local Universe. We present a homogeneous spatially resolved spectral analysis of a sample of nearby obscured AGN observed with deep ($>100$~ks) Chandra/ACIS exposures and all associated with extra-
nuclear extended emission. Exploiting the sub-arcsecond resolution of Chandra, we separate unresolved
nuclear emission from physically distinct extended regions and model their spectra using photoionized and thermal
plasma components. The extended emission is spectrally complex, requiring a combination of photoionized and collisionally ionized components in most regions. The spectral parameters exhibit systematic radial variations, with
nuclear regions typically hosting higher-column and higher-ionization components, while extended regions span similarly broad but overlapping ranges without a simple monotonic radial progression. Hard ($>3$ keV) X-ray emission in excess of
nuclear PSF spillover is detected in 19/20 sources, while extended Fe~K$\alpha$ emission is observed in 12/20 systems. These results establish a uniform baseline for the X-ray properties of heavily obscured AGN and provide constraints on the geometry and physical conditions of AGN-driven feedback in the local Universe.
Read the full abstract ▾
All authors R. Middei, G. Fabbiano, A. Trindade Falcão, D. Ł. Król, W. P. Maksym, M. Elvis
Databases
Intelligent Edge Computing
Kalgi Gandhi and Minal Bhise
The number of edge devices in large-scale edge systems is rapidly increasing. Edge devices have limited processing power,
memory, and network bandwidth, making resource utilization and data management during edge query processing challenging. Joins are among the costliest database operations in terms of time and resources. The State-of-the-Art edge query processing, Column Imprint-Hash Join CI-HJ, addresses this challenge using equi-height binning to accelerate hash joins. However, it lacks efficiency in real-time processing and scans unnecessary cachelines. This paper presents Workload Aware Column Imprint-Hash Join WACI-HJ, which uses a workload-aware approach to accelerate hash joins. Predicting the upcoming query workload in advance further improves its suitability for real-time edge query processing. WACI-HJ comprises two phases: WACI-HJ Generation Phase, including Pre-processing, Prediction, and Blocking and Hashing modules to compute bins based on the predicted workload before query arrival, and Query Processing and Resource Utilization, which handles query processing and CPU, RAM, and I/O utilization. Evaluations on a
benchmark dataset and a real-world Smart Transportation
dataset show a 54% reduction in cachelines read and 10% improved query execution time. The proposed technique is effective for both scaled and skewed data. Although PCR is an indirect measure of energy consumption, the work also directly measures energy consumption through energy-efficiency experiments. WACI-HJ shows 1%, 38%, and 49% gain in CPU, RAM, and I/O, respectively. Optimizing cache usage and query execution speeds up real-time traffic analysis, congestion management, and routing in Smart Transportation. Additionally, this technology can be applied to other domains to accelerate edge query processing.
Read the full abstract ▾
Artificial Intelligence
Asymmetries in Spontaneous and Instructed Deception
Josiah Luikham
Large
language models sometimes deceive users without being instructed to. However, much of the study on deception in models involves instructed deception. We investigated the relationship between instructed and spontaneous (uninstructed) deception in Llama-3.1-70B-Instruct. We compared these two deception settings through direction geometry, cross-setting classifiers, and cross-setting steering. We found the two deception settings share a component of direction (cosine of approximately 0.5) and an asymmetry in the transfer between settings regarding detection and causation. Spontaneous trained classifiers performed better on instructed data than vice versa, and instructed derived directions performed better at steering spontaneous
prompts than vice versa. Likewise the best
token position to derive steering vectors from differed from the best
token position to train and apply classifiers.
Read the full abstract ▾
Cosmology and Nongalactic Astrophysics
TDCOSMO XXVIII. The Hubble constant from the quadruply lensed quasar J1537$-$3010 with precise time delays
A. Galan, A. G. Schweinfurth, S. H. Suyu et al. (+21 more)
We present the first measurement of the Hubble constant ($H_0$) from the quadruply-lensed
quasar J1537$-$3010, which has optically-measured time delays at $\sim2 \%$ precision. We combine these delays with multi-band imaging data from the Hubble Space
Telescope (HST) and model the system with two independent software and teams. We adopt a mass profile that is maximally degenerate with $H_0$ to fully incorporate the mass-sheet degeneracy in the error budget, with nuisance parameters constrained by spatially resolved
stellar kinematics from the Multi Unit Spectroscopic Explorer (MUSE) and a line-of-sight (LoS) analysis using the Euclid Flagship simulation. The entire analysis is performed blindly to $H_0$, distances and mass density slope of the main deflector. After unblinding, we measure $H_0 = 75.5^{+9.3}_{-5.8}\ {\rm km\,s^{-1}\,Mpc^{-1}}$, corresponding to a $10\%$ precision measurement from a single system. This precision is driven by conservative lens modeling assumptions including differences between lens modeling methods and the limited
stellar kinematics constraints, from which we infer a total mass-sheet parameter $\lambda\equiv(1-\kappa_{\rm ext})\lambda_{\rm int} = 0.89^{+0.11}_{-0.06}$ ($\lambda_{\rm int}=0.89^{+0.09}_{-0.07}$) that is consistent with current results for elliptical
galaxies ($\lambda \approx 1$). Our
lensing constraints,
stellar kinematic measurements and LoS characterization will be included in subsequent population-level measurements of $H_0$. Moreover, our lens models combined with future near-infrared spectroscopy and imaging from the James Webb Space
Telescope will further reduce the uncertainties on $H_0$ from J1537$-$3010 alone, bringing it closer to the few percents precision of the time delays.
Read the full abstract ▾
All authors A. Galan, A. G. Schweinfurth, S. H. Suyu, D. P. Johnson, A. Chawla, D. Sluse, G. V. Kharchilava, E. Buckley-Geer, H. Lin, S. Knabel, W. Sheu, S. Ertl, T. Anguita, S. Birrer, F. Courbin, F. Dux, M. Millon, V. Motta, S. Schuldt, T. Treu, D. M. Williams, M. Cappellari, K. C. Wong, D. Eckert
Soft Condensed Matter
Polarisation-mediated underscreening from weakly bonded ion clusters
David Ribar, Jake W. Felber, Clifford E. Woodward et al. (+1 more)
We explore the hypothesis that ions form loosely connected clusters at high ionic strength in aqueous solutions, and that these clusters have relevance to the experimentally observed phenomenon usually referred to as "anomalous underscreening". Cluster formation lowers the ionic strength below its nominal value, slowing the decay of the screening length but not reversing it. Here we focus on an additional contribution, cluster polarisation. We demonstrate that this effect produces longer-ranged repulsive interactions between like-charged surfaces or particles in concentrated salt solutions. We derive an analytical bulk relation in which the entire architecture of a cluster enters through a single quantity, the charge-weighted second moment of its intramolecular charge structure factor, which quantifies the polarisation response of an arbitrary cluster
topology. We also make numerical calculations using classical
polymer Density Functional Theory, cDFT, for a model based on
star-like clusters in which satellite ions are weakly bonded by a harmonic spring to a common central ion. The corona of satellite ions is assumed to be net neutral, since the formation of highly charged clusters would be accompanied by a significant self-energy cost. Using this model, we calculate surface interactions and screening lengths at various overall salt concentrations. We show that, under the assumption that the fraction of ions belonging to clusters increases with salt concentration, one may qualitatively arrive at "anomalous underscreening", i.e., an effective screening length that displays a minimum at an overall (monovalent) salt concentration of about 1 M. Quantitatively, we note that the predicted growth of the screening length beyond this threshold value is weaker than typically found by experimental surface force measurements.
Read the full abstract ▾
All authors David Ribar, Jake W. Felber, Clifford E. Woodward, Jan Forsman
Computation and Language
Hamed Babaei Giglou, Jennifer D'Souza and Sören Auer
General-purpose NLP
embedding models perform well on linguistic tasks, but their ability to capture symbolic ontological structure remains unclear. We introduce AVA, a systematic framework for evaluating whether
embeddings distinguish logic-sensitive relational semantics in ontologies and knowledge graphs. AVA comprises 171,007 contrastive triplets derived from 163 heterogeneous ontologies using hierarchy inversion, relation substitution, and disjointness injection. Each triplet contains an ontology statement, a semantically equivalent paraphrase, and a logic-sensitive hard negative with contradictory relational meaning. We evaluate more than 25 state-of-the-art
embedding models and find substantial limitations: the best model achieves only 0.739 triplet accuracy, while hard negative accuracy falls to 0.135. Fine-tuning improves discrimination by a large margin but transfers poorly to downstream Semantic Web tasks, including taxonomy discovery and ontology
alignment. Further analysis suggests that improvements stem partly from perturbation-specific pattern recognition rather than robust ontological understanding. These findings reveal a persistent gap between linguistic representation learning and ontology-level discrimination, challenging the assumption that strong NLP
benchmark performance translates to Semantic Web competence.
Read the full abstract ▾
High Energy Astrophysical Phenomena
Portraits of the young and old in X-ray: New millisecond pulsar detections and updated characterization of young neutron stars
Wynn C. G. Ho, Patrick H. Wang and Sebastien Guillot
We present analyses of new and archival Chandra and NuSTAR data on two old millisecond
pulsars, a young highly-energetic
pulsar, and a young
neutron star. PSR J1810-0623 is a recently-discovered 4.55 ms
radio pulsar in a binary with a possible carbon-oxygen white
dwarf companion. We tentatively detect the
pulsar for the first time in X-ray using a new short Chandra HRC-I observation. Its faintness suggests a distance more in accord with the further (~1 kpc) of its dispersion measure-inferred distances. PSR J1933-6211 is a 3.54 ms
radio pulsar also with a carbon-oxygen white
dwarf companion. Our first-time X-ray detection in an archival HRC-S image indicates its X-ray luminosity is similar to that of the 1.9 Msun
pulsar PSR J1614-2230. Next, PSR J1813-1749 is a 1-2 kyr
pulsar with one of the highest known
spin-down luminosities. We measure its
spin period of 44.7 ms using 2025 NuSTAR and 2026 Chandra ACIS-S data and update its
spin evolution model. We show for the first time the pulse profile of PSR J1813-1749 at 3-79 keV, and we find that the spectra of the
pulsar and
pulsar wind nebula at these energies did not change significantly between observations in 2018 and 2025. Finally, for the 5-10 kyr
neutron star CXOU J182913.1-125113 in the
supernova remnant G18.95-1.1, we combine ACIS-I observations from 2009 and 2024 to better measure the
neutron star's spectrum.
Read the full abstract ▾
High Energy Astrophysical Phenomena
Evidence for a luminosity-class dichotomy in the supergiant donors of persistent high-mass X-ray binaries and supergiant fast X-ray transients
Lara Sidoli and Vito Sguera
We have investigated the properties of the supergiant donors in two types of high-mass X-ray binaries (HMXBs, i.e. the so-called supergiant fast X-ray transients (SFXTs) and the classical X-ray sources), by collecting all available information from the literature. A prevalence of O-type versus B-type supergiant donors in SFXTs was previously suggested in the literature, possibly indicative of faster winds in SFXTs than in classical, persistent sources. Our aim is to investigate whether there is evidence for a dichotomy in the
stellar luminosity class, since B-type supergiants are also known to be hosted in SFXTs. Our collection of the relevant literature allowed us to build a robust sample of both types of HMXBs with a secure
stellar classification. We have found that the SFXTs showing the most extreme, transient X-ray emission host only O-type or B-type
stars with a Ib and Iab luminosity class, whereas classical sources have mostly Ia donors. This is robust evidence that might explain the two different behaviours of HMXBs at X-rays since the
stellar luminosity class appears to be linked to the wind properties (the supergiant wind terminal velocity, the mass-loss rate, and the wind strength Q parameter), indicative of faster and less dense winds in SFXTs than in classical sources.
Read the full abstract ▾
High Energy Astrophysical Phenomena
GRMHD Simulations of Accreting Proto-Magnetars I. Implications for Gamma-Ray Burst Jets and Energetic Explosions
Tejas Prasanna, Kyle Parfrey, Brian D. Metzger et al. (+3 more)
Newly formed, rapidly rotating, strongly magnetized
neutron stars ("millisecond proto-magnetars") are promising central engines for gamma-ray bursts (GRBs) and luminous
supernovae. Although often modeled in isolation, they can be born surrounded by
accretion disks in
stellar collapse,
neutron-
star mergers, or
accretion-induced collapse. We present axisymmetric GRMHD simulations of hyperaccretion onto such objects, including a physical equation of state and charged-current weak interactions. Holding the weakly magnetized
accretion torus fixed, we vary the
stellar dipole field strength to span crushed-magnetosphere, magnetically channeled
accretion, and centrifugal-propeller regimes, and compare with an otherwise similar accreting black hole.
Accretion compresses the
stellar magnetosphere and opens additional
magnetic flux, producing relativistic jet powers that exceed isolated-dipole
spin-down estimates by factors of a few to ~10. Even while the magnetosphere remains compressed against the
stellar surface, stronger fields increasingly impede
accretion and enhance
outflows. Channeled-
accretion models show strong jet variability driven by plasmoid eruptions and intermittent magnetospheric
accretion, whereas the propeller model produces a steadier, more powerful jet and rapid
spin-down. The disk-magnetosphere interaction also regulates how efficiently the
neutron star grows and whether it
spins up or down; near
spin equilibrium, inefficient
accretion can delay collapse to a black hole relative to estimates based on the external mass-supply rate. Accreting proto-magnetars can therefore power relativistic jets and
baryon-rich
outflows with energetics comparable to those inferred for long GRBs and GRB-
supernovae. A companion paper explores implications for
neutron-rich ejecta and r-process nucleosynthesis.
Read the full abstract ▾
All authors Tejas Prasanna, Kyle Parfrey, Brian D. Metzger, Ore Gottlieb, Andrei Beloborodov, Koushik Chatterjee
High Energy Astrophysical Phenomena
Equation of state and neutron star properties with new mass-radius constraints from PSR~J1614–2230, PSR~J2124–3358, and 47~Tuc~X7
Melissa Mendes, Isak Svensson, Hannah Gottling et al. (+9 more)
We study the impact of new mass-radius information from PSR J1614-2230, PSR J2124-3358, and 47 Tuc X7 in a combined equation of state inference based on
chiral effective field theory constraints at
nuclear densities and using different high-density extensions, including perturbative QCD constraints. The largest impact stems from the heavy-mass PSR J1614-2230
star, which shifts heavy
neutron stars to smaller radii by around 0.4 km. Moreover, the combined astrophysical NICER, LIGO/Virgo, and X-ray information drives the radius posterior to a more data-driven distribution, which is less sensitive to the high-density extension. For the equation of state, the new mass-radius information significantly tightens the pressure and speed-of-sound posterior distributions, especially around three times saturation density. Finally, we make predictions for the poorly constrained masses of PSR J2124-3358 and 47 Tuc X7 based on the combined equation of state analysis and the other astrophysical sources.
Read the full abstract ▾
All authors Melissa Mendes, Isak Svensson, Hannah Gottling, Kai Hebeler, Achim Schwenk, Nathan Rutherford, Lucien Mauviard, Christine Kazantsev, Yves Kini, Denis Gonzalez-Caniulef, Sebastien Guillot, Anna Watts
High Energy Astrophysical Phenomena
A NICER view of PSR J1614$-$2230: a massive and compact millisecond pulsar
Lucien Mauviard, Sebastien Guillot, Lami Suleiman et al. (+17 more)
Using pulse profile modeling, we obtain the mass-radius measurement of a millisecond
pulsar (MSP) with data from the
Neutron Star Interior Composition ExploreR, XMM-Newton and the Chandra X-ray Observatory. We report here the radius of PSR J1614$-$2230, the second most massive MSP confirmed by
radio timing. All of the data sets are well described by a simple model composed of two circular hot spots. The final result yields an equatorial radius of $R_{\rm eq}=10.06 ^{+1.25}_{-0.87}\,$km, and a gravitational mass of $M=1.937^{+0.012}_{-0.013}\,M_{\odot}$ (equally tailed 68% credible intervals). Although a non-thermal component was previously reported at higher energies, we find no sign of it in either our phase-averaged or phase-resolved spectral analyses. Using new relations linking the compactness to oblateness or surface
gravity, and tailored to the
spin frequency of PSR J1614$-$2230, we infer a configuration with one hot spot near the pole, and another near the equator. The tight mass posterior is essentially informed by
radio timing, while the radius constraint is not as tight due to the low source signal (8.5$\sigma$ X-ray pulse significance). However, over all geometries and atmosphere models tested, the radius posterior tends toward low values ($\lesssim12.08\,$km, 90th percentile in all cases).
Read the full abstract ▾
All authors Lucien Mauviard, Sebastien Guillot, Lami Suleiman, Yves Kini, Denis González-Caniulef, Christine Kazantsev, Pierre Stammler, Devarshi Choudhury, Bas Dorsman, Mariska Hoogkamer, Daniela Huppenkothen, Tuomo Salmi, Anna L. Watts, Jérôme Novak, Natalie A. Webb, Jean-Francois Olive, Matthew Kerr, Lucas Guillemot, Ismaël Cognard, Gilles Theureau
Cryptography and Security
Explainable Artificial Intelligence for Industrial Cybersecurity: A Review of Methods, Operational Integration, and Research Challenges
Amr S. Mohamed, Charlotte Fritz, Ahmad Mohammad Saber et al. (+5 more)
The increasing digitalization of industrial infrastructure and the convergence of information technology (IT) and operational technology (OT) have expanded the cyberattack surface of industrial systems. To address the growing complexity of cyber threats, artificial intelligence (AI) and machine learning (ML) techniques are increasingly deployed within industrial cybersecurity operations, particularly in Security Operations Centers (SOCs). While these approaches improve anomaly detection, threat analysis, and automated response, their opaque decision-making presents challenges for operational trust, regulatory compliance, and incident response. EXplainable Artificial Intelligence (XAI) has emerged as a promising paradigm to improve the transparency and interpretability of AI-driven cybersecurity systems and decisions. This paper provides a comprehensive review of XAI techniques in industrial cybersecurity, focusing on industrial SOC environments and operational security workflows. We examine the role of AI in industrial SOC workflows, the types of operational data leveraged in industrial environments, and the benefits and limitations of AI-based threat detection. We then review major families of XAI approaches, including feature attribution methods, surrogate models, rule-based explanations, and visualization techniques, and analyze their applicability to industrial use cases. We further discuss the operational, regulatory, and safety requirements that distinguish industrial systems from traditional IT environments. Key challenges are examined, including limited labeled
datasets, model reliability, explainability-performance tradeoffs, and the integration of XAI tools into SOC workflows. Finally, we identify open research directions and opportunities for developing trustworthy, operationally viable, and domain-specific XAI-enabled cybersecurity solutions for industrial environments.
Read the full abstract ▾
All authors Amr S. Mohamed, Charlotte Fritz, Ahmad Mohammad Saber, Yiqun Ma, Ratinder Kaur, Mohammed Al-Darwbi, Daniela Friedrich, Deepa Kundur
High Energy Physics (Theory)
Higher anomalies, compressing SPTs and cohomology operations
Shani Nadir Meynet and Elias Riedel Gårding
We explicitly compute and tabulate the suspension map for
cohomology operations, applying it to the dimensional reduction of anomalies and higher group structures of generalised
symmetries in
quantum field theories and
lattice systems. In a
quantum system with a discrete $p$-form
symmetry, we can restrict the
symmetry defects to a codimension-$q$ subspace of spacetime, or equivalently confine the gauge field to a slab whose thickness approaches zero. This results in a $(p-q)$-form
symmetry on the subspace. We determine the fate of two important properties of
symmetries under this process. First, we compute the anomaly of the reduced
symmetry, or the higher anomaly of the original
symmetry, which acts as an obstruction to higher gauging and onsiteability. Second, for two
symmetries forming a higher-group structure, encoded as a Postnikov class, we compute the reduced higher-group structure, which may or may not be trivial. Both cases are captured by the $q$-fold iteration of the
cohomology suspension $\Omega$, also known as the loop functor or transgression, in the
cohomology of Eilenberg–MacLane spaces. Classical results state that $\Omega$ is an isomorphism in a stable range and annihilates mixed anomalies. To calculate $\Omega$ over a base ring which is not a field, we make use of small models for the chain DGAs of Eilenberg–MacLane spaces, pioneered by Cartan and Moore, whose results we review and extend. The critical added step is the improvement of a mod $p$ chain contraction to the $p$-local integers, obtained via a $p$-adic series. We automate the computations in a software package emcm, and provide extensive tables of the results. Finally, to make contact with explicit cochain formulations of
cohomology operations, we demonstrate how to calculate cup products and certain cup-$i$ products in this framework.
Read the full abstract ▾