Distributed Computing
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 3 September
Compound AI workflows are increasingly used to serve complex AI tasks by coordinating multiple AI models and software components. This approach enables deployment flexibility, as each workflow stage can expose different model variants and resource requirements, but it also expands the deployment choices. A deployment must choose an execution plan that selects AI models for each compound AI workflow stage and places them on a heterogeneous cluster in order to satisfy SLOs. Deployment optimizers therefore need estimates to compare many candidate plans and identify feasible ones. System metrics can often be profiled per stage and composed according to workflow topology, but accuracy cannot, as errors and information loss at upstream stages affect the accuracy of downstream stages. Existing approaches either profile complete configurations end to end, which scales poorly, or use product-based accuracy surrogates that treat stages as independent and can misrank candidate plans. We introduce Atlas, a framework for optimizing compound AI deployments under SLO constraints. Atlas uses MAP, a Markovian Accuracy Predictor, to estimate configuration accuracy from local conditional accuracy transitions between adjacent workflow stages. MAP discretizes intermediate outputs into accuracy buckets and composes transition profiles according to workflow topology, giving the optimizer an accuracy estimate without exhaustive end-to-end profiling. Atlas formulates execution-plan selection as a mixed-integer linear program that maximizes predicted accuracy subject to SLOs. Across four compound AI workflows, MAP achieves Spearman correlation up to 0.947 while reducing profiling cost by up to 2.6x relative to exhaustive end-to-end profiling. Guided by MAP, the Atlas optimizer selects execution plans within 0.03 of oracle accuracy while reducing deployment cost by up to 42% through heterogeneous placement.
Image and Video Processing
Explainable Multimodal Deep Learning Integrating Imaging and Clinical Data for Oral Potentially Malignant Disorder Detection
Oral potentially malignant disorders (OPMDs) are critical precursors to oral cancer, yet clinical detection remains challenging because of substantial phenotypic heterogeneity and overlap with benign conditions. Although image-based deep learning shows promise for automated screening, visual information alone may be insufficient in real-world settings, where diagnostic decisions also rely on patient-specific risk factors. We developed M2-OPMDNet, a multimodal deep learning framework that integrates co-registered white-light and autofluorescence intraoral images with structured clinical information for OPMD detection. A customized questionnaire was designed to capture clinically relevant risk factors and symptoms in a standardized, reproducible format for integration with image-derived features. Multiple image encoders, including conventional convolutional neural networks and foundation model-based architectures, were evaluated using a prospectively collected dataset reflecting real-world screening conditions. Model interpretability was assessed using SHapley Additive exPlanations (SHAP) to quantify feature- and modality-level contributions. M2-OPMDNet achieved an AUC of 0.952, outperforming unimodal approaches and showing improved performance for visually subtle lesions. SHAP analysis demonstrated that structured clinical variables contributed substantially to risk estimation and complemented imaging features. These results demonstrate that explainable multimodal learning combining white-light and autofluorescence imaging with structured clinical data can provide accurate, transparent, and clinically grounded OPMD detection. M2-OPMDNet offers a scalable framework for real-world oral cancer screening and decision support.
Read the full abstract
Computation and Language
LentEx: Generalizable Latent Entity Extraction via Synthetic Data and Instruction-Tuned LLMs
Latent entity extraction (LEE) tackles the challenge of identifying implicit, contextually inferred entities within free text-an area where traditional entity extraction methods fall short. In this paper, we introduce LentEx, a novel framework for latent entity extraction that leverages synthetic data generation and instruction fine-tuning to optimize smaller, efficient large language models (LLMs). Latent entities, which are often abstract and thematic, are crucial for applications such as retrieval-augmented generation (RAG), customer persona analysis, and knowledge graph enrichment. LentEx addresses the scarcity of labeled datasets by employing a template-based approach to generate diverse, contextually rich synthetic data, ensuring high variability and alignment with real-world distributions. To our knowledge, LentEx is the first to systematically approach LEE through the lens of LLMs. LentEx demonstrates significant performance improvements across multiple tasks, notably surpassing state-of-the-art models on the MTEB Clustering Benchmark. Furthermore, our methodology enables robust generalization to unseen domains, making LentEx highly applicable in real-world NLP tasks, including RAG and clustering, thereby establishing a new paradigm for latent entity understanding and extraction in natural language processing.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Implementation of frequency-correlated noise in CMB component separation: Method, Validation, and Early Applications
Parametric component separation methods used in cosmic microwave background (CMB) analyses commonly assume instrumental noise to be uncorrelated between frequency channels. This approximation may become insufficient for experiments whose design naturally introduces cross-channel correlations, such as interferometric or multi-band bolometric systems. This work aims to provide a consistent framework to include frequency-correlated noise in Commander2 and to quantify its impact on component separation in controlled and instrument-motivated simulations. We generalized the amplitude sampling of the Commander2 framework to consistently handle dense frequency-frequency noise covariance matrices, modifying the likelihood evaluation and parallel architecture. The implementation is validated on low-resolution simulations of a simplified CMB-S4-like configuration. We applied the generalized framework to instrument-motivated simulations of the QUBIC bolometric interferometer, using representative noise models derived from end-to-end simulations. Validation tests show excellent convergence and agreement between sampled parameters and analytical maximum-likelihood estimates. Accounting for frequency correlations reduces the width of the maximum-likelihood residual distributions by up to 8% on average, and up to 11% for individual parameters, relative to the diagonal-noise approximation. In QUBIC-based simulations, accounting for an anti-correlated noise produces posterior widths reduced by about 60%, while a positive-correlation test increases them by about 40%. This shows that neglecting cross-channel correlations can lead to incorrect posterior estimates in either direction. The generalized implementation presented here enables the Commander2 software to consistently propagate such correlations into parameter estimation, providing a framework for assessing their impact in future CMB analyses.
Read the full abstract
Databases
A Semantic Model of Genetic Evidence: A Step Toward Bridging the Basic-Science-Clinic Gap
Scientific and clinical decision-making depends on evidence from the primary literature, but existing standards for representing that evidence (FHIR Evidence, ECO, SEPIO, and the GA4GH Genomic Knowledge Standards) are oriented toward clinical-trial workflows, evidence codes, or single-variant assertions, and do not capture the fine-grained, domain-specific structure of claims in basic and pre-clinical research. We introduce a semantic model for scientific evidence with three core classes, specialize it for genetics, align it structurally to FHIR Evidence with a SEPIO-anchored credibility decomposition, and attach a compact dimensional vocabulary whose conditional-activation rules are validated by a SHACL schema for the implemented constraints. Using clinical variant interpretation as the driving use case, we evaluate the model through a human-AI annotation pilot over six genetics papers, yielding 28 evidence items and 95 source-anchored assertions, with a workflow that keeps curator-authored reference annotations distinct from AI-drafted annotations. Treating the pilot as a feasibility study rather than a benchmark, we argue that the model is a useful increment toward trustworthy, AI-ready infrastructure for variant interpretation: a reference data model and validation schema for representing genetic evidence.
Read the full abstract
Logic
Uncountably many maximally consistent neighborhood complete congruential modal logics
We solve an open problem posed by Peter Fritz in \cite{Fritz} by proving that there are uncountably many C-Post complete congruential modal logics which are neighborhood complete. The method is algebraic: we construct an uncountable sequence of varieties of modal algebras which are minimal in the lattice of all subvarieties of modal algebras and are generated by a single complete and atomic modal algebra.
Read the full abstract
Combinatorics
Degree-one Pieri rule for non-symmetric Jack polynomials
We give a degree-one Pieri formula for integral nonsymmetric Jack polynomials. In direct analogy with the symmetric case, its coefficients are described by a combinatorial device called jeu de fleches, a filling of composition diagrams. The proof is self-contained and uses reflection and raising recursions which appear to be new
Read the full abstract
Quantum Physics
State-Based Quantum Operations: Chameleon Gates
We introduce chameleon gates as a natural generalization of conventional quantum controlled-gates. Chameleon gates are state-based quantum controlled-operations that retain standard elements such as control and target systems, while introducing a new feature: the quantum knob. This knob is a quantum signal (state) that determines the operation performed by the gate. Consequently, the action and form of a chameleon gate depend dynamically on the quantum knob, allowing the gate to adapt its operation and implement transformations that are not necessarily unitary. This shapeshifting property is in stark contrast to conventional quantum controlled-gates, whose actions are fixed and cannot be modified. We also propose how chameleon gates can be realized using conventional quantum gates available in current quantum technologies. We then employ chameleon gates as a useful building block within the recently proposed state-based quantum computation (SBQC) framework. Using this approach, we demonstrate the simulation of state-dependent (nonlinear) quantum evolutions.
Read the full abstract
Data Structures and Algorithms
Statistics of Similarity Graphs in Node-Arrival Streams
In this paper, we study several statistical problems on similarity graphs in the node-arrival streaming model, including degree moments, diversity index, degree-moment sampling, and diversity sampling. We develop constant-pass, sublinear-space streaming algorithms for these problems and establish space lower bounds that nearly match the upper bounds in their dependence on the stream length.
Read the full abstract
Artificial Intelligence
BioSync: Transformer Fusion for a Multimodal Wearable Digital Biomarker
Multimodal wearable digital biomarkers must handle heterogeneous physiological signals, missing or corrupted sensor streams, and strict on-device memory budgets. We present BioSync, a compact transformer-based cross-modal fusion architecture (1,119 parameters; 4.6 kB in FP32 and 1.3 kB with INT8 weight quantization) designed for continuous wearable deployment. Across 11 synthetic physiological regimes and five real-world datasets (RESILIENT, AI-READI, WESAD, a stress-resilience HCI dataset, and PADS; N = 15 to 469), we evaluate BioSync alongside a quality-gated variant, a capacity-routed mixture, and classical and neural baselines. BioSync achieves the smallest worst-case regret across regimes (0.034 versus 0.048-0.107 for baselines) while remaining competitive on real-world wearable benchmarks. We also report mixed real-data findings for quality gating, test-time token masking, and capacity-routing weights, providing an empirical characterization of where compact transformer fusion helps and where simple baselines remain strong.
Read the full abstract
Quantum Physics
One-Shot and Concurrent Hitting Times for Grover-Coined Quantum Walks on Cubelike Graphs
We study the one-shot and concurrent hitting for the discrete-time Grover-coined quantum walk on cubelike graphs $G=\text{Cay}(\mathbb Z_2^d,\Omega)$ of degree $\Delta=|\Omega|$. Starting from the vertex labeled $0$, we identify $\sigma=\bigoplus_{\omega\in\Omega}\omega$ as a natural target vertex; for the hypercube, $\sigma$ is precisely the antipodal vertex. For families with $\Delta\to\infty$, let $T$ be an integer having the same parity as $\Delta$ and satisfying $ \left|T-\frac{\pi\Delta}{2}\right|\leq 1. $ We show that the probability $p_T(\sigma)$ of finding the walker at $\sigma$ when it is measured at time $T$ satisfies $$ p_T(\sigma)=1-O(\Delta^{-1/5}). $$ Thus the target is found with probability tending to one after $\Theta(\Delta)$ steps. For the concurrently measured walk, let $H_T^{\mathrm{Conc}}(\sigma)$ denote the probability that the target is detected at or before time $T$ when it is tested after every step. We prove $$ p_T(\sigma)\leq T H_T^{\mathrm{Conc}}(\sigma), $$ which implies $H_T^{\mathrm{Conc}}(\sigma)=\Omega(\Delta^{-1})$ over the same time scale. The proof uses the Walsh-Fourier decomposition, an exact two-dimensional reduction of each Fourier mode, and a universal second-moment identity for the associated character sums. Our results extend Kempe's hypercube hitting phenomenon (J. Kempe, Probab. Theory Relat. Fields 133, 215-235, 2005) to arbitrary cubelike generating sets and establish the conjectured asymptotic hitting behavior for cublelike and augmented cubes in Mulherkar, Rajdeepak and Sunitha (Int. J. Quantum Inf. 20,2250020, 2022)
Read the full abstract
Image and Video Processing
Fractional-Order Adaptive Motion Magnification: Phase-Reliability Weighting for Noise-Constrained Video Amplification
Eulerian video amplification boosts sub-pixel motion by band-pass filtering per-pixel intensity traces and applying a uniform gain. That gain ignores local structure, so sensor noise is amplified together with the signal, especially in textureless regions where the monogenic phase is unreliable. We propose FrAM (Fractional-order Adaptive Motion Magnification), a pipeline developed first offline and then as a causal stream. It replaces the constant temporal gain with a Grünwald–Letnikov derivative of fractional order, giving continuous control over high-frequency emphasis, and replaces the uniform spatial gain with a per-pixel weight derived from the local amplitude of the monogenic signal. On a controlled synthetic sequence split into textured and flat halves, FrAM matches the amplification of the Eulerian baseline while keeping flat-region temporal noise at the input level. The reduction holds across an eightfold range of input noise levels. Real videos show improved spatial selectivity and lower background noise in every case. The causal reformulation cuts the per-frame cost by two orders of magnitude, reaching 69\,fps at 640$\times$480.
Read the full abstract
Human-Computer Interaction
EyeMakeYou: Identity-, Task-, and Subjective-State-Conditioned Diffusion for High-Frequency Gaze Synthesis
Eye movement biometrics (EMB) is an emerging behavioral modality for user authentication, particularly in virtual- and augmented-reality systems, where gaze dynamics contain distinctive subject-specific features. However, robust EMB systems require diverse, high-quality gaze recordings that are expensive to collect and often unavailable at the scale needed for model development. Generative models can mitigate data scarcity, but existing methods either synthesize generic gaze behavior or personalize signals primarily by identity, without jointly representing the user's task and subjective state. Consequently, generated signals may appear visually realistic while failing to retain the behavioral properties required for biometric applications. To address this limitation, we propose EyeMakeYou, a multi-conditional denoising diffusion framework for subject-specific, high-frequency gaze synthesis. EyeMakeYou generates 5-s, 1000-Hz bivariate gaze-velocity sequences from an identity-removed reference trajectory and conditions the denoising process on an identity embedding, a task embedding, and self-reported ratings of overall difficulty, mental tiredness, and eye tiredness. Its objective combines diffusion noise prediction and identity preservation with multi-resolution spectral, drift-consistency, and event-weighted local-smoothness losses. Experiments on GazeBase show that EyeMakeYou achieves higher median spatial accuracy and greater real–synthetic similarity in the embedding feature space than the existing generative approaches, while retaining selected task-dependent associations between subjective reports and oculomotor features. These findings support conditional diffusion as a practical approach for augmenting gaze datasets for biometric and interactive applications.
Read the full abstract
Graphics
STyMo: Fast and Controllable Few-Shot Motion Style Transfer
Supporting a wide variety of motion styles is critical for creating diverse virtual characters, but current methods either require large stylized datasets or pre-trained models that cannot generalize beyond their training distribution. We present STyMo, a few-shot approach that learns motion style from only seconds of paired data and trains in one to two minutes. Our key insight is to decompose style into two components: a static component capturing time-invariant posture, and a temporal component capturing frame-wise dynamics. This decomposition yields an interpretable system where posture intensity, temporal exaggeration, and per-body-region style can be adjusted at runtime. Furthermore, the reduction in required training data and computation time structurally permits an iterative authoring workflow. To ensure robustness on arbitrary inputs, we further introduce a stylizability gate that automatically prevents artifacts on out-of-distribution motions. We demonstrate results across diverse motion styles, from subtle emotional variations to exaggerated character archetypes, and release our processed paired dataset to facilitate future research.
Read the full abstract
Geophysics
Low electrical conductivity of dry CaSiO3 perovskite under lower mantle conditions
Electrical conductivity (EC) provides important constraints on the composition and volatile distribution of Earth's deep mantle, yet the EC of davemaoite (CaSiO3 perovskite), a major lower-mantle phase, remains poorly constrained. We measured the EC of nominally dry CaSiO3 perovskite at pressures up to 89 GPa and temperatures up to 2200 K using impedance spectroscopy in a laser-heated diamond anvil cell. Conductivity increases with temperature but decreases systematically with pressure and it is substantially lower than previously reported, yet broadly consistent with recent theoretical predictions for oxygen-vacancy-mediated ionic transport. A distinct change in the temperature dependence coincides with the tetragonal-to-cubic phase boundary, revealing a modest enhancement of ionic transport across the structural transition. Along a normal lower-mantle geotherm, dry davemaoite is comparable in conductivity to bridgmanite near the top of the lower mantle but becomes progressively less conductive with depth. Under cold-slab conditions, dry davemaoite is substantially less conductive than dry subducted MORB and cannot account for the observed high-conductivity anomalies. Dry davemaoite therefore contributes little to bulk lower-mantle conductivity.
Read the full abstract
Nuclear Theory
Breakdown of the Plane-Wave Trojan Horse Analysis of the $^{12}\mathrm{C}+{}^{12}\mathrm{C}$ Fusion Reaction: Critical Role of Coulomb Distortions
Recently, a new Trojan Horse Method (THM) measurement of carbon-carbon fusion was reported by Li et al. [Phys. Lett. B (2026) 140675]. The purpose of the present work is to demonstrate the breakdown of the plane-wave approximation used in the analysis of these data and the critical role of Coulomb distortions in the initial and final states. The reaction mechanism underlying the THM analysis of the $^{12}\mathrm{C}+{}^{12}\mathrm{C}$ fusion reaction using the $^{16}\mathrm{O}+{}^{12}\mathrm{C}\to \alpha_s+\alpha+{}^{20}\mathrm{Ne}$ reaction is investigated. Particular attention is paid to the spectator momentum distribution and to the dependence of the THM reaction amplitude on the relative carbon-carbon energy $E$. It is demonstrated that agreement with the measured spectator momentum distribution does not by itself validate the plane-wave approximation. Although the experimental momentum distribution can be reproduced, inclusion of Coulomb distortions in both the initial and final channels leads to an energy dependence of the THM amplitude that is completely different from the plane-wave result. Consequently, the energy dependence of the $^{12}\mathrm{C}+{}^{12}\mathrm{C}$ fusion cross section extracted from the THM data can be strongly distorted by the plane-wave treatment. It is concluded that the astrophysical factor extracted in the plane-wave analysis cannot be regarded as reliable and may lead to misleading conclusions concerning the low-energy $^{12}\mathrm{C}+{}^{12}\mathrm{C}$ fusion reaction.
Read the full abstract
Game Theory
The Deterministic Hare Core Is Nonempty for Nine-Seat Approval Elections
Core stability gives approval-based committee elections a strong form of coalitional proportionality, but deterministic existence is not known for an arbitrary number of seats. We prove that every finite election with nine seats has a deterministic Hare-quota core committee. Starting from an eight-seat committee that is both a global Proportional Approval Voting (PAV) maximizer and core stable, we suppose that all one-candidate extensions are blocked and extract an inclusion-minimal closed response system. Singleton responses are eliminated by exact equality rigidity. For the remaining responses, rational Farkas certificates bound every nonpositive PAV add-marginal drift below by $-n/189$ and force every successor drift above $n/126$. The uniform response chain is irreducible, so stationarity makes these bounds incompatible. The computer-assisted component comprises an exact pointwise check over 7,356 voter types, 36 one-blocker cells, and a symmetry-complete family of 19 two-blocker motifs. Exact certificate archives and standard-library verifiers support these checks. The theorem settles the nine-seat case but does not decide deterministic core existence for arbitrary committee size.
Read the full abstract
Portfolio Management
Portfolio Diversification and Concentration under Dependence Uncertainty: A Majorization Approach
Modern portfolio theory identifies diversification as the primary tool for risk reduction. However, under model uncertainty, this cornerstone may no longer remain optimal. This paper investigates the tension between portfolio diversification and concentration under dependence uncertainty. In the absence of model uncertainty, we employ the framework of the majorization order and doubly stochastic matrices to formalize the degree of diversification, and prove that quasi-convexity is a necessary and sufficient property for a risk functional to be weakly consistent with the majorization order. We further derive worst-case risk measure inequalities and solve robust portfolio selection problems for a broad class of risk measures, including VaR, ES, Range-VaR (RVaR), and standard deviation (SD). Our results reveal a ”concentration paradox” for many widely-used risk functionals: when the dependence structure is fully ambiguous, robust optimization often recommends concentrating investment in a single asset to hedge against the worst-case dependence scenario. As an application, we propose a weighted robustness formulation that interpolates between a reference dependence structure and the worst-case structure. The formulation is structurally analogous to the constrained/unconstrained Expected Shortfall blend in the Fundamental Review of the Trading Book (FRTB) and provides a theoretical foundation for balancing diversification against robustness in the presence of model uncertainty.
Read the full abstract
Artificial Intelligence
Rethinking Indirect Prompt Injection as a Test-Time Search Problem
We formulate indirect prompt injection as a test-time search over a task-dependent attack surface induced by the environment, user task, and injection task. To operationalize this formulation, we introduce an agentic attacker with a dedicated search harness that performs environment reconnaissance, structured reasoning over attack strategies, and adaptive evaluation using victim-agent feedback. Across heterogeneous tasks, we find that increasing attacker test-time compute improves vulnerability discovery and exploitation, while ablations show that explicit strategy management is important for avoiding redundant search and sustaining gains at larger budgets. These results suggest that agentic security evaluations should characterize both the attacker's search procedure and compute budget, rather than treating attack success as a budget-independent property of the victim. More broadly, our findings identify the attacker's adaptive search over the system attack surfaces as an important and underexplored security risk for tool-using agents.
Read the full abstract
Machine Learning
Hakken: Predicting future discoveries to fill the gaps in today's knowledge
We present Hakken, a domain-agnostic prediction and explanation system performing knowledge prediction, i.e., growing scientific knowledge by establishing novel relationships, ones that are not limited to the deductive hull of previous knowledge. Hakken uses a transformer-based prediction model built on temporal sequences of knowledge graphs extracted from vast bodies of research publications, fused with an LLM's semantic knowledge, to predict the presence and define the type of as-yet undocumented relationships between scientific concepts. It then calls a model-agnostic explanation framework to provide accompanying information for each prediction that allows scientists to evaluate the suggested new relationship. While general purpose, we demonstrate Hakken's practical capabilities by applying it to the biomedical domain. There, Hakken's prediction model establishes a new benchmark for time-aware multi-label relation prediction, and we show that the model's output stays coherent and informative over extended time spans in historic data. In addition, we scored 1.5 million above-confidence-threshold hypotheses related to aging, qualitatively validated batches of these predictions with biologists and progressed three of them for empirical validation in wet-lab. Two predictions with potentially significant impact in the context of drug discovery and repurposing were confirmed, introducing previously undocumented interactions between TP53 and BAMBI, and between RAF1 and TNF, to biomedical science.
Read the full abstract
Artificial Intelligence
ResLearn-XR: Residual Learning for Network Traffic and Quality-of-Experience-Aware Modeling in Extended Reality
We present ResLearn-XR, a residual learning framework for predicting eXtended Reality (XR) network traffic and estimating Quality-of-Experience (QoE) risk. ResLearn-XR adopts a two-stage temporal learning structure comprising a base sequence prediction model augmented with task-specific residual learning components to improve adaptability to bursty, non-stationary XR traffic dynamics. The residual learning stages operate in the value space for continuous XR traffic forecasting and in the logit space for probabilistic QoE risk estimation. \rev{For the QoE-risk branch, we introduce a Data Descriptor Algorithm (DDA), a causal feature-construction module that converts packet-level application-layer observables into frame-timing-aware descriptors suitable for encrypted traffic analysis. We also construct an XR Traffic-QoE dataset that pairs continuous XR traffic traces with session-level user-reported QoE labels. ResLearn-XR reduces SMAPE by up to 17.84% across frame-count, frame-size, and inter-arrival-time prediction, while reducing QoE-risk estimation SMAPE by up to 87.8% over single-stage baselines.
Read the full abstract
Quantum Physics
Gradient-based optimal control of the non-Hermitian skin effect in optomechanical arrays
In single-port non-Hermitian sensors the Petermann factor offsets susceptibility gains, imposing a strict resource bound on metrological precision. We test whether a multi-port geometry can evade this bound: a double-chain optomechanical ladder with opposing non-reciprocal hoppings spatially separates signal amplification from quantum-noise drainage, and gradient-based differentiable optimal control (DOC) maximizes the resource-normalized Fisher information $\Fnorm$ subject to a Hurwitz-stability constraint. Across system sizes $N\in\{\num{6},\dots,\num{16}\}$ the optimizer returns $\Fnorm>0$ in every case, with two coexisting solution classes whose selection is initialization-dependent: deep-stability configurations achieve $\Fnorm\in\numrange{0.937}{0.987}$ with attenuated transmission, while marginal-stability configurations deliver directional gain $\Gfwd\in\qtyrange{13.5}{15.5}{\dB}$ with isolation $\Iso\in\qtyrange{40}{64}{\dB}$. A multi-restart ensemble reveals these classes are the endpoints of a precision–gain frontier. All solutions remain Hurwitz-stable under \qty{5}{\percent} disorder (\qty{87.5}{\percent} recovery), and the deep-stability advantage survives realistic preamplifier noise at $\Fnormeff\approx\num{0.3}$–$\num{0.5}$. Mapped onto circuit-QED parameters, the architecture enables sub-attonewton force sensing and broadband axion searches across the \qtyrange{1}{10}{\giga\hertz} band.
Read the full abstract
High Energy Astrophysical Phenomena
Preferential accretion onto eccentric and unequal binary black holes
Supermassive binary black holes (SMBBHs) are expected to be surrounded by circumbinary disks (CBDs) which affect the binary through gravitational forces and accretion. It has been reported that the binary can experience “preferential accretion” where one black hole (BH) out-accretes the other for hundreds of orbits, but this asymmetry has yet to be fully described or understood. In this work, we utilize a suite of 80 SMBBH hydrodynamical simulations with varying mass ratios ($q_b$) and eccentricity ($e_b$) in order to robustly delineate the behavior of preferential accretion, determine its relationship to the structure of the CBD, and study its observational consequences. We characterize the accretion-rate ratio $\lambda(t) \equiv \dot{M}_2(t)/\dot{M}_1(t)$ and the mass-ratio rate of change $\dot{q}_b \equiv d/dt(M_2/M_1)$ across the suite. We confirm that the secondary tends to out-accrete the primary ($\lambda \geq 1$), and find this preference to be strongest for low-$e_b$, low-$q_b$ binaries and increasingly time-variable toward high $e_b$. We also find that (i) the time-variability of $\lambda$ tracks the precession of the CBD, (ii) there can be sub- and super-Eddington accretion in a single binary, and (iii) the gas-driven approach toward equal mass becomes particularly slow for highly eccentric, high $q_b$ binaries, suggesting that some binaries may not reach $q_b=1$ within the $30\,\mathrm{Myr}$ lifetime of a quasar and therefore allowing LISA to constrain the accretion history of SMBBHs. Our findings also suggest that periodically flickering jets are a potential observable signature of many binaries.
Read the full abstract
Artificial Intelligence
When Quantization Breaks Memory: Recurrent-State Write-Back in Low-Precision Temporal Inference
Quantization is widely used to reduce the computational and memory demands of neural-network inference. In recurrent networks, however, the quantized state is stored and returned at the next time step, so the rule used to store that state can alter subsequent computations. Here, we introduce recurrent-state write-back to denote this rule and isolate its effect in a compact GRU encoder–decoder for fluorescence lifetime imaging, a molecular imaging modality used in quantitative biological imaging. A central task is estimating two lifetime parameters, the short-lived component {\tau}1 and the long-lived component {\tau}2, from high-noise time-resolved fluorescence signals. Holding the trained model fixed, replacing continuous state propagation with deterministic 4-bit state storage increases estimation errors for {\tau}1 and {\tau}2 by approximately 70x and 300x, respectively. Failure occurs when repeated small updates remain below the write threshold, leaving the stored state nearly fixed while the network continues to propose change. Error feedback, residual memory, and direction memory carry information from these suppressed updates across time and recover accuracy without retraining. Precision sweeps show that increasing state precision can worsen a fixed recurrent solution, while matched training shows that compatibility with the state interface can be learned. To test whether this behavior extends beyond the GRU, we repeat the post-training intervention in an independently trained LSTM, where coarse write-back reproduces the failure, error feedback restores accuracy, and state-specific interventions reveal greater sensitivity of the cell state than the hidden state. Our results establish recurrent-state write-back as a key determinant of low-precision recurrent dynamics and identify the state-storage interface as a central design consideration for quantized recurrent inference.
Read the full abstract
Computation and Language
Towards Understanding Pause Token Fine-Tuning Dynamics: A Mode Retention Perspective
Pause-token methods improve LLM reasoning by inserting special tokens into sequences. Prior work explains these gains through computational expressivity. However, there is relatively little investigation into the training dynamics of pause tokens. We explore how pause tokens reshape the training dynamics of fine-tuning. Two controlled pilots expose distinct asymmetries. On a synthetic continual-learning task, masked pauses overwrite a previously-learned distribution roughly 4x less at matched final adaptation (H1, mode retention); on a synthetic math-reasoning probe, the boundary-adjacent token comes to encode substantially more downstream-step information (H2, non-myopic compression). We formalize a training rule consistent with both - Masked Boundary Pause (MBP), pause tokens placed at reasoning-step boundaries with their loss masked. Across 1B-8B Qwen and Llama models, MBP consistently improves reasoning, achieving gains of up to 6 points on math and 2.5 points on code, while preserving general language understanding abilities. We further demonstrate that this mode-preserving strategy extend gains to GRPO. These results recast pause tokens as a training-dynamics intervention on the retention-adaptation trade-off, rather than merely an inference-time computation device.
Read the full abstract
Cosmology and Nongalactic Astrophysics
The Super-Sample Covariance of Line-Intensity Mapping Power Spectrum
In this work, we provide the first derivation of the line-intensity mapping (LIM) power spectrum super-sample covariance (SSC) from first principles, and also derive as a by-product the non-Gaussian in-box contributions to the covariance for the first time. Previous studies have typically modelled the LIM power spectrum covariance using either the Gaussian approximation or estimates obtained from mocks or the data itself, neglecting uncertainties related to whether the limited volume surveyed sits in a cosmological overdensity. This contribution, known as the SSC or, depending on the context, the field-to-field variance, cannot be estimated from the data, but it is crucial for a correct inference of global quantities, i.e., for ensemble-averaged parameters rather than the actual values just within the patch of the Universe observed. For our derivation, we employ a combination of the halo model and standard perturbation theory that allows us to capture the nonlinearity and non-Gaussianity of the covariance. After a successful validation of our predictions against painted N-body simulations, we explore different scenarios related to current and future LIM experiments, quantifying the relative importance of the non-Gaussian in-box and SSC. We find that the newly derived contributions to the LIM power spectrum covariance are crucial at intermediate and small scales, especially for cases in which the covariance is not dominated by instrumental noise. We find that the relative relevance of the SSC with respect to the other covariance contributions is roughly independent of the survey volume, but does depend on the specific response of the power spectrum to large-scale modes for each line and redshift. Therefore, the impact of the SSC will be increasingly significant for parameter inference from the current and the next generation high signal-to-noise LIM surveys.
Read the full abstract
Systems and Control
Estimating the Health and State of Charge of Each Cell in a Second-Life Battery System from Field Data
Effective use of battery storage depends on reliable estimation of its state of health (SOH) and state of charge (SOC). Model-based state estimation requires the open-circuit voltage (OCV) curve, which is typically unknown for second-life batteries. We present a framework that jointly estimates the states and parameters of an equivalent circuit model solely from field operation data, using Gaussian process regression to reconstruct the OCV curve. Applied to a real second-life battery system of 27 modules and 324 cells, it reveals SOH heterogeneity, a systematic SOC imbalance, and two faulty cells, all validated against a reference measurement. We aggregate the cell SOH and SOC to module level and benchmark them against a lumped-module model fitted without the individual cell voltages. The lumped-module model follows the average behavior and cannot capture the limiting cells, overestimating SOH by up to 31% and SOC by up to 23%.
Read the full abstract
Networking
Uncertainty Signals for Network Intent Translation: Risk Ranking and Ambiguity Localization
Intent-based networking realization starts by translating high-level intents into low-level network configurations. Recent approaches have shifted toward LLM-based translation. Despite promising results, most studies focus on translation accuracy and overlook risks associated with deploying the resulting configurations. In this work, we investigate the pre-deployment translation risk of LLM-generated configurations by analyzing the model's uncertainty. We propose to use two uncertainty signals, namely sampling-based predictive uncertainty for translation-risk ranking and token-level entropy for ambiguity-source localization. We evaluate these signals on an ambiguity-controlled test set across different context types and sampling budgets, using a Llama-3.1-8B-Instruct model fine-tuned for intent translation on a vendor-specific switch platform (Juniper EX3300). The results demonstrate that predictive uncertainty provides a useful signal for ranking translations by risk across context types and sampling budgets, albeit with substantial miscalibration under less informative contexts. Moreover, we show that parameter-token entropy correlates with parameter-sourced ambiguity and keyword-token entropy correlates with description-sourced ambiguity. These results indicate the potential of using uncertainty signals in an LLM-generated configuration deployment pipeline, where predictive uncertainty can support selective deployment, while token-level entropy can identify sources of ambiguity.
Read the full abstract
Computation and Language
Cultural Misalignment in Large Language Models: Detection, Measurement, and Mitigation Through Targeted Fine-Tuning
We evaluate three open-weight LLMs (Gemma3-12B from the USA, Bielik-11B-v3 from Poland, and Qwen3-4B from China) against World Values Survey Wave 7 data for 63 demographic personas across three countries, using normalized Wasserstein distance to quantify distributional misalignment. Contrary to expectations, no model favors its home country: the Chinese-built Qwen3-4B performs worst on its own Chinese population (W1 = 0.436, the highest misalignment in the entire model x country matrix). Targeted LoRA fine-tuning on the five worst-case personas, requiring fewer than 1,200 training pairs and under 15 minutes on a single GPU, reduces bias by 16.8% for Bielik-11B (p_Bonf = 0.002, d = -4.4) with all five targets improving. However, country-level decomposition reveals that fine-tuning redistributes rather than removes bias: Bielik's worst-case personas swap entirely from American to Chinese elderly, with zero overlap between pre- and post-correction sets. To our knowledge, this is the first study to target worst-case demographic personas with LoRA fine-tuning for cross-cultural bias mitigation.
Read the full abstract
Computation and Language
Patterns of Priming in Production: Lexical, Semantic and Structural Alignment in Language Model Generation
This paper investigates structural priming in language model (LM) production, examining how preceding structural context influences sentence completion. While prior work has demonstrated priming effects in comprehension of structural alternations, it remained unclear whether these persist in production, where, when generating, an LM samples from many possible continuations at each step. We address this question through a series of controlled sentence-completion experiments on dative constructions. In line with prior work, we find that LMs are susceptible to structural priming, particularly in sentences that are semantically coherent. In terms of priming magnitude, we find that while there is a greater relative increase of double-object datives against our baselines, in line with inverse frequency effects, there is a larger absolute increase in prepositional-objects, the more frequently produced construction. Finally, we not only observe that structural priming is boosted by lexico-semantic coherence, but that structurally primed completions display greater levels of lexico-semantic repetition. Taken together, our evidence supports the view that structural priming in LMs operates across multiple levels of linguistic representation, facilitating, and facilitated by syntactic, lexical, and semantic alignment. Code: https://github.com/the-context-lab/primedproduction.