Peer Revenue Publish or profit

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Thursday 3 September

Optimization and Control

Data-Driven Brownian Reflection Control

We study a data-driven reflection control problem for a Brownian model with unknown drift and volatility. We first propose a learn-then-optimize (LTO) algorithm: it estimates the policy-relevant parameter during exploration, plugs the estimate into the optimality equation, and exploits the resulting policy—achieving an $O(\sqrt{T})$ finite-time expected regret bound. We further propose two algorithms, adaptive-updating (AU) and full-history adaptive-updating (AU-FH), which continuously update the estimator and reflecting level, attaining an improved $O(\log T)$ regret bound. Notably, AU-FH algorithms leverages all historical data, yielding better performance in numerical simulations. Our analysis decomposes regret into exploration, transient, and learning components. Transient regret from nonstationarity is bounded by the time-integrated deviation of the transition semigroup from stationarity evaluated on the holding cost, which can be further bounded via a Foster-Lyapunov inequality for exponential convergence of the controlled reflected Brownian motion (RBM). For learning regret, we establish local regularity properties together with consistency and mean-squared error bounds for the estimator, which control the stationary cost gap between the learned and optimal reflection policies. In addition, we leverage the monotonicity of the moving boundary Skorokhod map to derive moment bounds for the AU algorithms' switching states, via pathwise comparison with fixed-boundary RBMs.
Read the full abstract

Analysis of PDEs

Sharp asymptotics for a transport model with a nonlocal condition of the Fisher-KPP type at the boundary

This paper is concerned with the precise asymptotics, as time goes to infinity, of a transport problem in a half plane coupled with a nonlinear nonlocal boundary condition. This system arises from a class of models for the spatial spread of epdemics, its space independent version being the classical Kermack-McKendrick model. Using ideas pertaining to the study of nonlocal equations of the Fisher-KPP type, and exploiting the particular structure of the model, we prove that any initially localized solution will lag behind the minimal traveling wave, with a delay that grows logarithmically in time.
Read the full abstract

Human-Computer Interaction

GazeFS: Target-Centered Gaze-Trajectory Forecasting and Stabilization from Gaze-Head History

Target-centered gaze interaction requires more than suppressing frame-to-frame fluctuations: target acquisition produces task-aligned changes in gaze-head dynamics, while a gaze trace may retain a persistent target-relative residual direction. We formulate gaze correction as online target-centered gaze-trajectory forecasting and stabilization and introduce GazeFS, which maps a variable-length gaze-head history to the next target-center direction and a short-horizon Search/Focus estimate without target information at inference. Across 7,960 acquisition episodes from 30 participants, Search-Focus differences remain stable under quality control, onset exclusion, and duration matching. History windows improve phase decoding over the current endpoint, but explicit task progress remains a strong control. Under the 30-participant, five-fold grouped out-of-fold protocol across three seeds, the reductions relative to raw hold in Focus episode bias, within-episode dispersion, and P90 target error are 0.182 degrees, 0.257 degrees, and 0.400 degrees, with participant-bootstrap 95% confidence intervals excluding zero. Endpoint-free replay from empty history preserves the Focus advantage and yields raw-network phase balanced accuracy/AUPRC of 0.925/0.993; coordinate controls further show that recent history contributes beyond explicit progress metadata. GazeFS therefore improves Focus target centering and empirical residual contraction while leaving temporal smoothness as a separate objective.
Read the full abstract

All authors Yaozheng Xia, Zaiping Zhu, Bo Pang, Minghao Xie, Hui Li, Shaorong Wang, Sheng Li

Superconductivity

Tunable Chern superconductivity of PtBi$_2$ in slab geometry

Recent experiments indicate that PtBi$_2$ is a Weyl semimetal whose surfaces become topological superconductors with i-wave gap symmetry at low temperature. The bulk hosts 12 Weyl cones while in the superconducting state each of the two surfaces hosts in addition 6 Majorana cones. We study a simplified model in a slab geometry, where the finite thickness gaps out the bulk Weyl cones and leaves the surface Majorana cones as the low-energy degrees of freedom. We identify two competing mechanisms that gap these cones: inter-cone hybridization opens a trivial gap, whereas a Zeeman field opens a nontrivial one. In the nontrivial regime, chiral Majorana edge modes connect the gapped cones. We determine their wavefunction profile and relate it to the momentum-space distribution of the cones. Mapping out the phase diagram, we find both trivial and nontrivial regions in the quasi-two-dimensional limit.
Read the full abstract

Instrumentation and Methods for Astrophysics

nufftcf: Fast Auto- and Cross-Correlation Function Estimation for Irregularly-Sampled Time Series via the Non-Uniform FFT

Estimating the auto-correlation function (ACF) and cross-correlation function (CCF) of sampled series is a standard task in many physics fields, including astronomy and environmental sciences. We present nufftcf, an ACF/CCF estimator designed primarily for irregularly sampled series, numerically consistent with established kernel-weighted definitions and scaling as O(nK) for K requested lags, rather than O(n^2). nufftcf was initially developed to evaluate the Wiener-Khinchin theorem for irregularly sampled data using the Non-Uniform Fast Fourier Transform (NUFFT), through the Flatiron Institute FINUFFT library, with Gaussian and rectangle kernels. An O(n)-per-lag two-pointer scan replaces the naive O(n^2) computation of the effective pair count used to normalize each lag bin. This enabled real-space estimators using the same kernels, providing exact reference implementations with the same O(nK) complexity and, in some cases, lower cost than the NUFFT path. The library also provides dedicated classical-FFT estimators for regularly sampled data. All estimators share a common calling convention. Using synthetic time series, we validate nufftcf for ACF against pastas, a library used in groundwater time-series analysis, and for CCF against pyZDCF, used in astronomical time-series analysis. Benchmarks confirm O(nK) scaling for the NUFFT and real-space nufftcf estimators, compared with O(n^2) for the pastas slotting technique. nufftcf is already advantageous at moderate series lengths because of its low millisecond-scale overhead. For regularly sampled data, the FFT path further reduces computational cost. We also demonstrate nufftcf on a simulated ground-based stellar light curve combining quasi-periodic rotation, correlated flicker noise, seasonal sampling gaps, and heteroscedastic measurement errors. Notebooks and scripts are provided to reproduce the examples and explore other use cases.
Read the full abstract

Biological Physics

A regenerating free-energy register for protein-templated period-2 DNA synthesis by Drt3b

Deng et al. discovered that Drt3b synthesizes protein-primed poly(AC) DNA without a nucleic-acid template. The structures reveal a finite protein architecture and product-associated contacts, but not the renewable dynamical rule that converts a bounded pocket into a long period-2 sequence. We propose that Drt3b is a regenerating protein-substrate free-energy register. The framework specifies a sampled quantum mechanics/molecular mechanics (QM/MM) potential of mean force and a preregistered, within-domain first-order electronic-descriptor map for edge-specific activation barriers; physical curvature is admitted only through a training-frozen Taylor-remainder bound that widens uncertainty but never corrects a held-out mean. Conditional hazards derived from those barriers are averaged over unresolved conformational, hydration, protonation, and metal-coordination trajectories to yield a semi-Markov event kernel. Pointwise ratios of competing hazards are therefore fixed by the same barrier differences and preregistered prefactor rules that govern exit timing, so nucleotide choice and dwell statistics are co-generated rather than separately fitted. A regime gate separates stationary or slowly driven experiments, analyzed by a joint sequence-time cycle determinant, from genuinely nonstationary protocols, analyzed by the full age-structured tilted propagator. The decisive test is one locked parameterization: without condition-specific refitting, it must jointly predict nucleotide choice, waiting-time laws, product-length tails, and the rank, minors, null spaces, and rescue structure of perturbation responses. An independent 2026 DRT3 study supplies a stringent cross-construct transport test after sequence and structural alignment are frozen. The proposal requires neither coherent quantum computation nor coupling to an external field.
Read the full abstract

Quantum Physics

Five-Dimensional Compatibility and Stratified Local-Unitary Structure of Pure Three-Qubit States

We investigates the joint compatibility problem of the three pairwise concurrences, the three-tangle, and the Kempe invariant in arbitrary pure three-qubit states, and gives necessary and sufficient conditions for these five quantities to be simultaneously realizable by the same pure state. This five-dimensional compatibility region can be viewed as a fiber structure over the previously derived four-dimensional reachable region. We further find that the residual continuous local-unitary freedom depends on the position of the state within this region. In the generic interior, one continuous degree of freedom remains. It is fixed if the three-tangle vanishes or a pairwise concurrence becomes zero, and also at the boundary of the four-dimensional tangle region. The Kempe invariant gives the simplest algebraic parametrization of the fifth coordinate, but the same degree of freedom may be described by a known pure-state entanglement monotone. The compatibility conditions can also be expressed in terms of concurrence-based bipartite measures and multipartite quantities fixed by the pairwise concurrences and the three-tangle.
Read the full abstract

Mathematical Physics

Classification of abstract Bose field models

We classify a class of abstract Bose field models in quantum field theory up to unitary equivalence. The class includes abstract free Bose field models, abstract van Hove–Miyatake models, and infrared-renormalized van Hove–Miyatake models. Moreover, as an application of our classification, we classify quadratic interaction models. In particular, our results answer a question of A. Arai concerning the classification of infrared-renormalized van Hove–Miyatake models.
Read the full abstract

Statistical Mechanics

Equivalence classes of finite-time transitions in optimal control and non-equilibrium relaxation

We present a theory for the optimal control of stochastic systems in structured environments, represented by penalty terms in the cost functional. We show that such control problems generically feature sharp finite-time transitions associated with a qualitative change in the control strategy at a critical time. Starting from an overdamped Langevin equation and a quadratic cost functional, we show that all resulting problems fall into three canonical equivalence classes (parabolic, hyperbolic, and elliptic), distinguished by the sign of the determinant of the control Hamiltonian. For each class, we obtain the optimal protocol, the cost function, and the critical time in closed form, and show that the transition exhibits features of a continuous phase transition at mean-field level. We then establish a mapping between the optimal control cost and the large-deviation rate function governing non-equilibrium relaxation after a potential quench. The mapping covers the parabolic and hyperbolic classes, while the elliptic class has no simple relaxation counterpart. This correspondence implies that recently discovered finite-time dynamical phase transitions, which are exponentially costly to sample directly, are accessible through ordinary averages over optimally controlled trajectories. To validate our theoretical findings, we report three experiments with optically trapped colloidal particles: a control transition for the mean stochastic work, and the finite-time dynamical phase transitions in free diffusion and in harmonic relaxation.
Read the full abstract

All authors Jan Meibohm, Samuel Monter, Clemens Bechinger, Sarah A. M. Loos

Software Engineering

No One Left Behind: Cross-Level Analysis for Sustainable Software Engineering

Software engineers expect a software system to be efficient, maintainable, economically viable, and socially responsible throughout its lifecycle. Unfortunately, decisions made at the organizational, process, or product levels of software development - even when they improve one of these goals - often have unintended long-term consequences at other levels. In response, software-engineering research has explored different ways to achieve software sustainability, including energy efficiency, resource optimization, and code maintainability. However, existing research offers limited explanations for how sustainability problems emerge and reinforce one another across the socio-technical levels of software engineering. Moreover, we argue that sustainability challenges are not isolated but rather systemic: They stem from interactions among organizational priorities, development practices, and technical conditions. In this vision paper, we introduce the concept of sustainability antipatterns to capture conditions that systematically produce unsustainable outcomes. We demonstrate how this concept can help expose cross-level misalignments between socio-technical levels and sustainability dimensions that remain difficult to recognize through artifact-centric or dimension-specific analyses. To guide future research, we outline concrete directions for identifying, formalizing, detecting, and mitigating sustainability anti-patterns. We envision sustainability anti-patterns as a foundation for an integrated and actionable research agenda on sustainable software engineering.
Read the full abstract

Artificial Intelligence

Adapting to Evolving Requirements: Agentic AI for Retail Supply Chain Operations

Retail supply chain operations rely on coupled decision modules that must adapt as requirements evolve. LLMs offer a natural-language interface for this task, but existing methods primarily focus on individual optimization models. Extending them to heterogeneous decision pipelines is challenging because a requirement may admit multiple intervention paths with different downstream effects. We formulate requirement-driven adaptation as the joint selection of an intervention route and an admissible module-level change, and propose a graph-constrained agentic framework in which domain agents expose admissible reformulation interfaces and a central processor searches over bounded intervention paths. Candidates are validated and compared using downstream KPIs. In collaboration with a large retail partner, we evaluate 100 warehouse requirements elicited from practitioner interviews, with GPT, Qwen, and DeepSeek as base LLMs. Relative to direct LLM reformulation, our framework improves correctness and end-to-end success across all three models, raising end-to-end success from 72–76% to 79–83%.
Read the full abstract

All authors Lei Zheng, Liping Yang, Zihao Li, Guodong Lyu, Chaik Ming Koh, Chung-Piaw Teo

High Energy Astrophysical Phenomena

A catalogue of insights from the fourth LIGO-Virgo-KAGRA observing run

Over the last decade, gravitational-wave science has developed from an era where a single detection was a Nobel-winning achievement to an (almost) everyday occurrence. The LIGO-Virgo-KAGRA (LVK) Collaboration's flagship scientific results are now their catalogues of detections, which include a comprehensive set of analysis results characterising the astrophysical properties of sources, and their implications for fundamental physics and cosmology. We review discoveries from the fourth observing run (O4) of the LVK network. The most recent catalogues each include $\sim100$ new detections. Among these new detections are remarkable single discoveries: the loudest signals detected so far, and sources unlike those seen before. However, significant advances in the field are coming from population-level analyses, leveraging the increasing sample size to infer constraints on the astrophysical distribution of merging compact binaries, potential deviations to general relativity, and the expansion rate of the Universe. Analyses of the latest data have revealed evidence for multiple subpopulations of sources. These observations provide a unique probe into the evolution of massive stars and the formation of coalescing binaries. Our understanding of gravitational-wave signals and their sources will continue to deepen with more observations. The final catalogue from O4}is due for release in December 2026, and future observing runs with even greater detector sensitivity are planned. These will bring even larger catalogues, and additional loud signals, continuing to advance gravitational-wave science.
Read the full abstract

Machine Learning

High-Dimensional Learning Dynamics of Attention-Indexed Models

Attention mechanisms are central to modern foundation models, yet their training dynamics remain poorly understood, especially when the attention matrices have extensive rank. In this work, we study attention-indexed models, a broad framework that can represent multi-layer and multi-head attention architectures. First, we show that, in a suitable high-dimensional limit, the population-loss landscape is characterized by a finite set of trace order parameters. In contrast, online stochastic gradient descent (SGD) is governed by an infinite hierarchy of matrix moments, which we show can be exponentially well-approximated by a finite truncated system. Second, this framework reveals that attention parameterization itself can act as an architectural implicit bias. Direct optimization of an attention matrix $S\in\mathbb{R}^{d\times d}$ can remain trapped in an uninformative state. Tied attention ($S=WW^\top$) induces an automatic symmetry-breaking mechanism and yields weak recovery in $\Theta(d^2\log d)$ samples. For untied attention, $S=UV^\top$, we uncover a fast-slow mechanism: the pre-activation mean first evolves on a fast timescale, while the overlaps evolve on a slower one. Weak recovery on the $\Theta(d^2\log d)$ scale occurs when the state selected by the fast dynamics breaks the initial symmetry.
Read the full abstract

All authors Yizhou Xu, Margarita Sagitova, Lenka Zdeborová, Florent Krzakala

Information Retrieval

GRASP: Graph-Retrieval Automated Scoring Pipeline for Label-Free Multi-Topic Essay Grading

Automated short-answer grading research has historically focused on exams consisting solely of questions pertaining to a single topic. Automatic grading of exams containing questions about more than one topic remains less explored. In this work, a Graph-Retrieval Automated Scoring Pipeline (GRASP) is introduced for grading label-free multi-topic science exams. Label-free exams are short-answer exams in which a student's responses to several distinct topics are merged into a single paragraph, with no markup labels or segmentation indicating which span answers which question. Reference answers for each question are encoded into a FAISS vector index via Sentence-BERT, and a semantic similarity graph is constructed over this set of reference answers. At grading time, sentence count heuristics, with a large language model used to resolve ambiguous cases, are first applied to predict how many distinct topics were answered in the student essay. This process is performed without training data or domain-specific example essays. Candidate reference nodes, each storing one (question, reference answer, concatenation of both) from the reference index, are then retrieved through cosine similarity based Retrieval-Augmented Generation (RAG) and Graph Retrieval-Augmented Generation (GRAG). GRAG operates by taking the top cosine matches as seed nodes and then performing a graph traversal over strong edges to find additional reference nodes that may have been missed by RAG. The Hungarian algorithm is then used to optimally assign one reference node per question segment such that no reference is duplicated. Each segment is then graded against its assigned reference independently using GPT-4.1-mini. This experiment is performed to show the effect of retrieval quality on grading accuracy and the benefit of graph-augmented retrieval versus strict cosine similarity methods at various levels of essay complexity.
Read the full abstract

Probability

Consensus time for asynchronous $\ell^p$ relaxation: graph dependence

We study the asynchronous $\ell^p$ relaxation introduced by Amir, Nazarov, and Peres: at each step, a uniformly chosen vertex minimizes its incident $\ell^p$ energy. For the profile $f_t$ after $t$ updates, let $$ \mathsf T_p(G,1/2):= \sup_{\lVert f_0\rVert_\infty\le1} \mathbb{E}\bigl[\min\{t\ge0:\operatorname{osc}(f_t)\le1/2\}\bigr]. $$ For $1<p<\infty$, we obtain graph-dependent estimates for boxes, trees, and conductance expanders. On the nearest-neighbor box $[L]^d$, for $d,L\ge2$ and $n=L^d$, the answer is, up to logarithmic factors, $nd^{1/(p-1)}L^{p/(p-1)}$ for $1<p<2$ and $ndL^2$ for $p\ge2$. On bounded-degree trees, an explicit rerooting-invariant parameter $T_G$ determines the answer up to logarithmic factors; for arbitrary trees it gives upper and lower bounds that differ additionally by the maximum degree. If the volume conductance $h(G)\ge h_0>0$, then $\mathsf T_p(G,1/2)=\Theta_{p,h_0}(n\log n)$ without a degree assumption. At $p=\infty$, every connected graph satisfies $\mathsf T_\infty(G,1/2)\ge c nD^2/\Delta$, where $D$ and $\Delta$ are its diameter and maximum degree.
Read the full abstract

Functional Analysis

A Fejér–Riesz inequality for Dirichlet series

We prove the following inequality for Dirichlet polynomials: \[ \int_0^1 |f(1/2+\sigma)|\,d\sigma\lesssim \lim_{T\to\infty} \frac{1}{2T} \int_{-T}^T |f(it)| \, dt. \] In particular, for a Dirichlet series $f(s) = \sum_{n\geq 1} a_n n^{-s}$ belonging to the Hardy space $\mathscr{H}^1$ of Dirichlet series, \[ \left|a_1+\sum_{n=2}^\infty \frac{a_n}{\sqrt n\log n}\right| \lesssim \|f\|_{\mathscr{H}^1}. \] This answers a question raised previously in the literature and it proves that the multiplicative Hilbert matrix has a bounded symbol.
Read the full abstract

Strongly Correlated Electrons

Universal Driven Critical Dynamics of Entanglement Entropy

The Kibble-Zurek mechanism (KZM) and finite-time scaling (FTS) provide a foundational framework for driven critical dynamics, yet their predictive power has been largely confined to local observables. Here, we establish a universal finite-time scaling theory for the nonequilibrium dynamics of quantum entanglement. Using unbiased quantum Monte Carlo simulations, we investigate the corner entanglement entropy of (2+1)-dimensional interacting Dirac fermions driven from ordered phases toward a quantum critical point. We find that the corner entanglement accurately obeys a universal driven scaling governed by the driving rate and system size, persisting whether the initial ordered state is fully gapped or hosts gapless Goldstone modes. Crucially, this dynamical entanglement exhibits a logarithmic dependence on the driving rate, from which the universal corner coefficient of the underlying conformal field theory can be robustly extracted far from equilibrium. These results generalize the KZM from local observables to the intrinsic nonlocal quantum information measures, offering a practical blueprint for characterizing quantum criticality and entanglement on programmable quantum simulators.
Read the full abstract

Computers and Society

Bridging Formal and Perceived Fairness: Development of an Interdisciplinary Framework in Algorithmic Decision-Making

While fairness has become a central concern in research on algorithmic systems, the field remains predominantly shaped by Computer Science, resulting in a strong emphasis on formal fairness metrics and bias mitigation strategies. Nevertheless, this focus may obscure a fundamental challenge: fairness is not merely a technical property, but a subjective, context-sensitive human judgment shaped by cognitive heuristics, mental models, normative expectations, and sociotechnical factors. Crucially, users' perceptions of fairness may diverge substantially from the fairness criteria an algorithm formally satisfies; a system may meet predefined technical fairness requirements yet still be perceived as unjust by decision-affected stakeholders. In such cases, the system fails on a fundamental dimension: it will not be trusted, accepted, or considered legitimate. Taking a user-centered design perspective, this paper presents a work-in-progress conceptual framework that bridges Computer Science approaches to formal algorithmic fairness with normative and Social Science fairness approaches regarding perceived fairness, trust, and technology acceptance, embedding both within the sociotechnical conditions that shape human judgment. Through (1) theoretical literature synthesis, (2) interdisciplinary workshops, and (3) stakeholder interviews, the project aims to inform evaluation approaches that integrate computational fairness audits with user-centered assessments and guide the design of fairness-aware, human-centered algorithmic systems that support informed, well-calibrated fairness judgments by those affected.
Read the full abstract

All authors Maike Lindermayr, Mattia Cerrato, Luisa Hübner, Johannes Kraus

Disordered Systems and Neural Networks

Design Principles for Reproducible Networks

From protein complexes to electronic circuits, many natural and engineered systems function only if assembled in an exact, reproducible fashion. The structure of each of these systems can be understood as a network, yet network science lacks the mechanisms to consistently reproduce exact topologies, focusing instead on generating network ensembles. We introduce the framework of network design where we encode the local constraints obeyed by a system's building blocks in a design set, and derive the Unigraphical Design Theorem, which determines when these constraints guarantee reproducible assembly into a unique structure, a process we call unigraphical assembly. For systems whose design sets do not specify a unique outcome, we identify guided assembly as a second route to reproducibility, in which temporal ordering decomposes construction into unigraphical steps. Applying these results to 3,618 reproducible systems, including protein complexes, molecules, and robots, we classify those that undergo unigraphical assembly and those that require guided assembly. We further identify a diversity-redundancy boundary that explains how systems trade component variety for structurally interchangeable parts while retaining unique assembly. Finally, we experimentally test the theory using 3D-printed components to re-engineer generative construction sets into systems that assemble unigraphically into prescribed topologies. Network design thus reframes reproducibility as a mathematically testable property of real networks, opening a route to the rational engineering of complex systems.
Read the full abstract

Machine Learning

Pushing the (Decision) Boundaries: Dynamically Calibrating Differentially Private Noise to Explainability in Federated Learning

Federated Learning (FL) with Differential Privacy (DP) is increasingly adopted to preserve data confidentiality in distributed machine learning. However, DP noise distorts learned representations and degrades explanation fidelity, limiting differentially private FL where trustworthy explanations are required, such as assistive clinical diagnosis. Prior work adapted DP noise with static feature-importance signals, restricting explainability to post hoc analysis and precluding noise calibration to explanation quality during training. We propose XCal-FL, a closed-loop, explainability-driven local training algorithm for image classification in cross-silo FL that dynamically calibrates DP noise from three complementary signals: (1) prediction logit variations, measuring causal influence on model confidence, (2) counterfactual margins, capturing decision-boundary sensitivity, and (3) saliency concentration, quantifying spatial coherence of model attention, while enforcing formal DP guarantees via adaptive privacy accounting. Experiments on three medical imaging datasets across varying FL configurations show that XCal-FL yields more accurate and interpretable global models, improving predictive performance by over 10% and explanation fidelity by up to 5$\times$ over static-noise FL, and outperforming state-of-the-art adaptive DP methods in fidelity. XCal-FL also achieves higher privacy-budget efficiency, turning each unit of cumulative privacy loss into larger gains in both accuracy and explanation fidelity. Our analysis further reveals that, unlike predictive performance, which scales roughly linearly with privacy loss, explanation fidelity exhibits non-linear dynamics. These findings suggest explainability is a distinct dimension of the privacy trade-off that cannot be inferred from utility alone, with implications for training and privacy-budget allocation in decision-critical applications.
Read the full abstract

All authors Michael Khavkin, Kichang Lee, Jaeho Jin, JeongGil Ko, Eran Toch

Optics

Thermal Noise in Optical Coatings: A Rigorous Opto-Mechanical Comparison of Nanolaminates and Isotropic Mixtures

We investigate whether a perfectly ordered periodic nanolaminate is intrinsically less mechanically dissipative than an isotropic co-sputtered mixture designed to provide the exact same optical function. Focusing on a room-temperature SiO2/TiO2 system, we compare coating Brownian dissipation under strict parity of optical thickness and effective refractive index. Amorphous phases are modeled using complex Young's moduli and real Poisson's ratios. The comparison is highly non-trivial: while the nanolaminate is optically optimal, its structural anisotropy forces in-plane shear to operate at the highly dissipative Voigt upper bound, mechanically penalizing it against an isotropic medium. The nanolaminate is treated via exact optical (Wiener) and mechanical (Backus) relations. The isotropic mixture is described by optical closures (Lorentz–Lorenz, Bruggeman) coupled to a complex mechanical self-consistent scheme. To assess uncertainty from the unknown co-sputtered morphology, we explore rigorous optical and mechanical admissible regions in the complex plane. Owing to the small intrinsic losses of SiO2/TiO2, these regions collapse into extremely narrow lenses, strictly limiting microstructural uncertainty. Across all realistic isotropic closures, the nanolaminate systematically exhibits lower mechanical dissipation than the co-sputtered mixture. Even against the thermodynamic lower bound of the isotropic state, this topological advantage remains valid for high effective refractive indices. We prove the ordered architecture exploits the optical upper bound to minimize the required volume fraction of the highly dissipative high-index phase, overwhelmingly overcompensating for its intrinsic mechanical anisotropy penalty.
Read the full abstract

All authors Vincenzo Pierro, Vincenzo Fiumara, Veronica Granata, Guerino Avallone

Cryptography and Security

NACRE: Rethinking Confidential Containers through Native Architectural Support

Linux containers achieve high density and fast lifecycle operations by sharing the host kernel, but this design also lets a compromised host inspect or modify container state. Existing confidential-computing systems protect an enclave address space or an entire guest operating system, while recent container-granularity systems still add a separate protection context. These abstractions do not make a dynamic group of host-managed Linux processes the architectural protection unit. This paper presents NACRE, a RISC-V hardware-software co-design for native confidential containers. Its key insight is to separate the host's authority to manage resources from its authority to access or commit protected state. Hardware-recognized container identities direct protected traps to an isolated S-mode agent, while an M-mode monitor commits security- sensitive identity, mapping, and page transitions. The agent delegates services to host Linux without changing satp; services that neither access private bytes nor modify protected state also avoid M-mode. We prototype NACRE by extending QEMU, OpenSBI, Linux, a trusted agent, and runc. The prototype implements the single-container private-memory substrate and covered launch, fault, fork/COW, user-access, and teardown paths. Across five lmbench syscall and pipe metrics, the three-run means remain within 3.5% of the runc-origin baseline. With the eight nginx object-size means weighted equally, aggregate throughput is 1.9% lower.
Read the full abstract

All authors Linke Song, Wenhao Wang, Weijie Liu, Rui Hou

Analysis of PDEs

The mean-field limit of the Schrödinger-Lohe model and emergent dynamics

The Schrödinger-Lohe (SL) model is a coupled system of nonlinear Schrödinger equations describing the temporal-spatial evolution of the component wave functions, and it corresponds to the infinite-dimensional counterpart of the Lohe matrix model for quantum synchronization. In this paper, we study a rigorous mean-field limit of the SL model and provide a quantitative estimate on the fluctuations between one-marginal distribution and one particle distribution in 2-Wasserstein distance. For the derived kinetic SL equation, we present sufficient conditions leading to the complete and practical synchronizations.
Read the full abstract

Analysis of PDEs

A counterexample to McKean's conjecture for the Landau-Coulomb equation

We disprove McKean's conjecture, which asserts that the entropy dissipation is monotone nonincreasing along solutions, or equivalently that the entropy is convex in time, for the spatially homogeneous Landau-Coulomb equation. We provide an explicit counterexample which consists of a Maxwellian equilibrium under radially symmetric, smooth perturbation on an annulus of scale $R\gg 1$. Developing the derivative of the entropy dissipation in powers of $R$ reveals that the leading order terms can be positive for a suitably chosen perturbation. Remarkably, the counterexamples are close to equilibrium in relative entropy and standard Sobolev norms. Since the Landau equation arises from the Boltzmann equation in the grazing collisions limit, our counterexample also shows that McKean's conjecture is not true in general for the Boltzmann equation.
Read the full abstract

Game Theory

EF1-Constrained Nash Social Welfare with Identical Additive Valuations: Complexity, Guarantees, and Experiments

We study the allocation of indivisible goods among agents with identical additive valuations, focusing on envy-freeness up to one good (EF1) and Nash social welfare (NSW). Since every maximum-NSW allocation is EF1 under additive valuations, the associated threshold problem inherits the known strong NP-hardness of NSW maximization under identical additive valuations and is strongly NP-complete. We therefore focus on welfare guarantees satisfied by arbitrary EF1 allocations. Although every such allocation is known to achieve an $e^{-1/e}$-approximation to the unrestricted optimal NSW, we identify conditions yielding stronger guarantees. Under uniform valuations, every EF1 allocation is NSW-optimal. Under an $\varepsilon$-small-item condition, every EF1 allocation achieves an explicit approximation ratio $\rho_n(\varepsilon)$ satisfying $\rho_n(\varepsilon) = 1-O(\varepsilon^2)$ as $\varepsilon\to 0$ for fixed $n$. We further consider the stronger sequential requirement that $\operatorname{EF1}$ be maintained after every item assignment. For this setting, we introduce PriorityNet, a deep reinforcement learning framework trained with Proximal Policy Optimization (PPO) and equipped with prospective $\operatorname{EF1}$ action masking, which guarantees prefix-wise $\operatorname{EF1}$ by construction. Across 3,000 test instances in each of the offline full-information and random-order online regimes ($n\in[2,20]$, $m\in[5,100]$), PriorityNet achieves mean normalized $\operatorname{NSW}$ values of $0.9911$ and $0.9701$, respectively. Relative to the offline Longest Processing Time (LPT) heuristic and the online least-valued-bundle rule, it attains instance-wise win-minus-loss rates of $+27.10\%$ and $+17.87\%$. Its aggregate welfare matches the offline LPT baseline to four decimal places and modestly improves upon the online baseline, from $0.9694$ to $0.9701$.
Read the full abstract

All authors Zih-Sian Yang, Yi-Hao Chen, Yu-Te Kuan, Cheng-Jui Wu, Chuang-Chieh Lin, Po-An Chen

Complex Variables

Projective embedding of degenerating family of Kähler-Einstein manifolds with normal crossing limit

We study the Bergman embeddings of degenerating families of Kähler-Einstein manifolds of negative curvature. We show that when the central fiber has only simple normal crossing singularities, we can construct orthonormal bases so that the induced Bergman embeddings converge to the Bergman embedding of the limit space together with multi-bubbles.
Read the full abstract

Cryptography and Security

Flip, Don't Shuffle: Watermarking LLMs at the Speed of Inference

We introduce Stateless Bernoulli Watermarking (SBW), a new statistical watermark for Large Language Models that determines green list membership through independent per-token Bernoulli trials. Unlike KGW's vocabulary permutation or SynthID's multi-layer tournament, SBW requires only a single comparison per token against a counter-based random number generator, reducing membership complexity to $O(1)$ and enabling single-kernel execution with zero intermediate allocations. We prove that this formulation preserves the same detection guarantees as fixed-size green lists: the z-score test remains $\mathcal{N}(0,1)$ under the null. The stateless architecture enables capabilities unavailable to existing methods: full-vocabulary self-salt watermarking (over 6000$\times$ faster than KGW's self-salt and 2$\times$ faster than SynthID despite biasing the entire vocabulary with candidate-dependent seeding) and architectural compatibility with distributed inference. In end-to-end generation benchmarks, SBW adds less than 1% overhead at all batch sizes. We additionally identify hash function design as a previously unexplored axis for watermark quality, showing that a GPU-native Jenkins hash improves null calibration by 1.8$\times$ while producing more diverse text. Experiments across two seeding schemes and eight $(\gamma, \delta)$ configurations confirm statistical equivalence with ROC-AUC differences below 0.01.
Read the full abstract

Systems and Control

Can Julia land on the Moon? On the development of a GNC simulation framework for the Argonaut lunar lander

No, the Julia programming language cannot land on the Moon - but it can play a crucial role in designing and analysing the Guidance, Navigation, and Control (GNC) algorithms required for doing so. This paper presents the development of a lunar landing simulation framework implemented in Julia at the European Space Agency (ESA), within the Argonaut lunar lander programme. ATLAS (Argonaut Tools for Landing Analysis and Simulation) is a modular suite of analysis and simulation tools that cover the complete descent and landing phase of Argonaut, integrating high fidelity translational and rotational dynamics, varying mass properties, propellant sloshing, detailed sensor and actuator models, and flight-representative GNC algorithms within a multi-rate simulation environment. The framework is intended to bridge early-phase prototyping and large-scale Monte Carlo analysis within a single environment. This work evaluates the advantages and limitations of adopting Julia compared to established GNC development practices based on the MATLAB/Simulink ecosystem. The results show that Julia provides a powerful, flexible, and high-performance environment for agency-driven research, early-phase design studies, and computationally intensive closed-loop simulations enabling large-scale, parallelizable simulations and rapid design iteration cycles.
Read the full abstract

Machine Learning

Is One Step Enough for Offline Policy Improvement?

Behavior regularization in offline reinforcement learning limits the exploitation of critic errors, but strong anchoring can also restrict policy improvement. We study how policy improvement is composed through multi-step proximal policy improvement (MPI), which re-centers each proximal objective on the preceding policy. We parameterize the procedure by a nominal total horizon $T$ and $K$ stages with local horizon $T/K$, distinguishing subdivision at a fixed total horizon from additional refinement at a common local horizon. Our analysis shows that sequential re-centering can reach endpoints unavailable to any single proximal step and characterizes how subdivision reduces the leading local discretization error of ideal updates under a fixed critic. We consider TD3+BC and IQL-based policy extraction to examine how improvement composition interacts with actor objectives and policy geometry. TD3+BC experiments on D4RL locomotion suggest that subdivision can broaden the range of useful total horizons, while adding refinement stages at a fixed small local horizon can improve return. The results identify improvement composition as a design choice alongside regularization strength, with distinct effects from horizon subdivision and additional policy extraction.
Read the full abstract

Mesoscale and Nanoscale Physics

The Casimir free energy of peptide films on a silicon substrate: Impact of dielectric-to-metal transition in silicon and nanoparticles in peptide

Using the Lifshitz theory of the van der Waals and Casimir forces, we calculate the Casimir free energy of thin peptide films deposited on silicon substrates. The Casimir free energy is found as a function of film thickness for different fractions of water in the film, in the presence of either nonmagnetic or magnetic nanoparticles, and under the impact of irradiation of a silicon substrate with laser pulses or dopants resulting in the dielectric-to-metal phase transition. It is shown that for a dielectric silicon there is the borderline value of the film thickness, such that the Casimir free energy is negative and contributes to the film stability for thicker films, but is positive and makes the film less stable for thinner ones. According to our results, the borderline value of peptide film thickness decreases with increasing volume fractions of water and in the film. This decrease is more pronounced for the magnetic nanoparticles and becomes stronger with increasing their radius. The borderline value of peptide film thickness is found as a function of the fraction of water in the film. If the silicon substrate is in metallic state, the Casimir free energy of peptide coating is always positive, which makes it less stable. Possible applications of the obtained results in organic electronics and biomedicine are discussed.
Read the full abstract