Machine Learning
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 4 September
The Appropriately Combined Edge-length (ACE) sequence in A-BLiN depends on the zooming dimension $d_z$. This note removes that dependence. The next edge length is selected from the number of cubes that survive the preceding elimination. The resulting Count-Adaptive BLiN algorithm does not use $d_z$ or the zooming constant $C_z$, yet it attains $\widetilde{\mathcal O}_d(T^{(d_z+1)/(d_z+2)})$ regret with $\mathcal O_d(\log\log T)$ batches. Together with the adaptive-grid lower bound in Theorem 10 of the original paper, the optimal batch complexity remains $\Theta_d(\log\log T)$ when $d_z$ is unknown.
Statistics Applications
Symptom clusters in Long COVID in the UK: prospective community-based cohort study using unsupervised machine learning
Long COVID is a condition usually defined by persisting symptoms following infection by the SARS-CoV-2 virus beyond the acute phase of infection. The condition has a significant impact on healthcare systems, the economy, and the individuals living with it. Due to the diverse and extensive symptomatology of long COVID, symptom co-occurrence tracking can be used to capture patient experiences and improve understanding, diagnosis, and management. Here, we leverage the UK Office for National Statistics (ONS) COVID-19 Infection Survey (CIS). The CIS was run between April 2020 and March 2023. On February 3, 2021, the ONS launched a new CIS question evaluating self-reported symptom persistence of 23 long COVID symptoms post self-reported COVID-19 infection. We use Jaccard symptom-by symptom distance matrices, derived from the binary survey responses of the presence of each symptom. Three methods are then used to visualise symptom co-occurrence: heatmaps, non-metric multidimensional scaling, and agglomerative hierarchical clustering with complete linkage. We split our analysis into two parts, first looking at all long COVID survey responses (n = 207,319) and then looking at responses stratified by time-since-onset of long COVID up to 24 months (n = 30,224 at zero months-since-onset). We find higher symptom co-occurrence in prolonged long COVID. We also find clusters of neurological/systemic symptoms, gastrointestinal symptoms, and respiratory symptoms, with separability in symptom co-occurrence by organ system becoming less pronounced as time-since-onset of long COVID increases. Shortness of breath, weakness/tiredness, and muscle ache present as core symptoms long COVID. We demonstrate that the symptom experience of long COVID evolves from early stages to late stages with higher symptom burden and multi-systemic presentation. We also find three possible phenotypes of early long COVID.
Read the full abstract
Machine Learning (Statistics)
PAC-Bayesian Reconstruction Guarantees for Time Series Variational Autoencoders
Forecasting time series accurately is critical for applications with complex data ranging from energy systems to healthcare and finance. Among current state of the art models, generative latent variable models are increasingly implemented; yet principled generalisation guarantees for modern latent variable models remain limited. In particular, while Variational AutoEncoders are widely used for sequential data, their theoretical analysis is largely restricted to i.i.d. settings. In this work, we develop a PAC-Bayesian framework for latent variables models applied to time series. Building on reconstruction-based bounds, we extend PAC-Bayesian guarantees to Markovian latent structures, capturing temporal dependencies through a sequential generative process. These guarantees do not grow with the length of the trajectory. Our bounds depend on assumptions which are common in the literature; we provide an example framework where they would be verified to show that they are not as restrictive as they may seem.
Read the full abstract
High Energy Physics (Theory)
CFT characters from localization
We explain a general procedure to compute characters of chiral algebras of 2d (S)CFTs geometrically from their coadjoint orbits, in analogy with the coadjoint orbit method for finite-dimensional Lie groups. We demonstrate that this method can be used to give concise rederivations of the characters of the Virasoro algebra, Kac-Moody algebras, as well as supersymmetric generalizations.
Read the full abstract
Computer Vision
BLASt3R: Bundle Adjustment of Any Image Set with Multi-View Matching and Monocular Priors
Recent hybrid Structure-from-Motion (SfM) systems combine the robustness of feed-forward 3D reconstruction with the accuracy of traditional bundle adjustment (BA) with pixel matching. They are usually the best performing methods however their scalability and usability remains limited since estimating dense correspondences between views is prohibitively costly, especially considering time constraints inherent to online applications like Visual SLAM (VSLAM). In this paper, we introduce a regularized BA framework that leverages a fast multi-view matcher and monocular priors for initialization and regularization. In contrast to existing systems, our unified approach seamlessly supports both online VSLAM and offline reconstruction from unordered image collections within the same optimization framework and sharing common hyperparameters for all tasks. Extensive experiments across both domains demonstrate improved performance and speed tradeoffs over traditional, feed-forward, and hybrid baselines. Notably for VSLAM, our uncalibrated method outperforms all previous calibrated approaches.
Read the full abstract
High Energy Astrophysical Phenomena
Exploring the connection between Fast Radio Bursts and binary neutron star mergers
Fast Radio Bursts (FRBs) are highly energetic radio sources whose duration is of the order of milliseconds. The physical origin of these sources is still unknown. Many models suggest magnetars as possible progenitors of FRBs, and this is supported by the association between FRBs and the Galactic magnetar SGR 1935+2154; other proposed progenitors include binary neutron star (BNS) mergers, that are themselves linked to magnetar formation. In this work we investigate the possible connection between FRBs and BNS mergers, including magnetars that might be produced in such events, by comparing the detection rates inferred from synthetic BNS and associated FRB populations with the rates observed by CHIME. We produce a synthetic catalog of BNS mergers by combining recent theoretically predicted BNS merger rate as a function of redshift and the neutron star mass distribution inferred from measurements of Galactic BNSs. Using this catalog we predict the number of BNS systems ending as magnetars (stable or supramassive neutron star) or black holes (formed promptly or after the collapse of a hypermassive neutron star) for different equations of state. We then simulate for each BNS (and therefore for each magnetar remnant) an associated FRB and estimate how many of them can be potentially detected by CHIME. We find that the rate of BNS mergers and the rate of magnetars produced after BNS represents a non-negligible fraction of the FRBs detected by CHIME, both repeating and non-repeating. Although additional formation channels need to be considered to account for the entire population of FRBs, the existence of a fraction of FRBs that might genuinely be associated to BNS mergers has profound implications in the context of multi- messenger astronomy, supporting the systematic searches of coincident detections of FRBs and GWs from a BNS merger with current and future facilities.
Read the full abstract
Instrumentation and Methods for Astrophysics
Active focal plane phase mask coronagraphy with a discrete pixelated device: I. Study of the theoretical performance trade space
Recent advances in high-contrast exoplanet imaging instrumentation have introduced the concept of adaptive coronagraphy. For example, liquid-crystal-on-silicon spatial light modulators can be used as programmable phase masks or digital micro-mirror devices as configurable pupil apodizers. Adaptive coronagraphy offers the ability to adjust in real time to changing observing conditions and science goals, such as switching between blind surveys and follow-ups of known objects, or optimizing observations of multiple star systems including binaries and triples. At the same time, these active devices present challenges: finite spatial sampling, limited phase resolution, and the scalar nature of their modulation can all reduce coronagraphic performance. We look at the performance of a coronagraph utilizing pixelated discrete focal plane masks in function of various key parameters, notably spatial sampling, phase resolution, temporal jitter, and calibration errors. The analysis includes several FPM designs: vortex, four-quadrant phase mask, Roddier and Roddier, dual-zone phase mask, and azimuthal cosine phase mask. Both monochromatic and 20% broadband imaging conditions are considered, along with the absence or presence of a central obstruction from a secondary mirror in the telescope pupil, associated with variations in Lyot stop sizing. Our results identify the main error sources and limiting parameters of scalar, pixelated FPM coronagraphs. Under ideal conditions, performance is mainly limited by spatial sampling and chromaticity of scalar phase modulation. These devices are therefore primarily relevant to ground-based high-contrast imaging rather than the deeper contrast regimes required by future space observatories. Although motivated mainly by SLM implementations, the conclusions should also apply to other discrete pixelated focal-plane phase masks.
Read the full abstract
Machine Learning (Statistics)
FluxDisco: Symbolic Regression for Stoichiometric Dynamical Systems via Monte Carlo Graph Search
Dynamical symbolic regression methods identify governing differential equations from noisy data, balancing interpretability and predictive accuracy. However, standard methods often produce expressions that violate known physical laws. To address this, we propose FluxDisco, a physics-informed framework tailored for flux-based, stoichiometric ODE systems. By leveraging a known stoichiometry, we reduce the expression search space and ensure physical adherence. Our framework adapts the Monte Carlo Graph Search algorithm for the unique challenges associated with joint flux discovery of stoichiometric systems. We evaluate our method across a range of physical and biological systems, demonstrating its ability to accurately recover governing dynamics through interpretable equations.
Read the full abstract
Robotics
Morphology and actuation as inductive biases in robotic hand manipulation
Robotic hands vary widely in anatomical fidelity and mechanical complexity, and these structural choices influence the coordination of joint motions and the difficulty of controlling the system. A unified framework is presented in which the kinematic and actuation stages are analysed separately and in composition, through the conditioning of the task Jacobian, the actuation matrix, and their product. It is applied to two hands representing opposing design philosophies, the Shadow Dexterous Hand and the Anatomically Correct, Biomechatronic Hand, along four morphological aspects: joint axis geometry, actuator-to-DOF ratio, coupling architecture, and authority distribution. All parameters are derived from the hands' canonical digital representations. Anatomical fidelity carries no uniform advantage: oblique axes improve thumb conditioning but leave the long fingers worse conditioned than the orthogonal-axis design, while the branching tendon network improves the effective control mapping at every long finger and worsens it significantly at the thumb, where actuator authority is concentrated on thumb opposition. Predictions derived from these metrics are evaluated against reinforcement learning experiments using PPO, DDPG+HER, and TQC+HER, across three different tasks.
Read the full abstract
Instrumentation and Methods for Astrophysics
Laboratory characterization of Hierarchical Fringe Tracking
One of the main limitations in long-baseline interferometry lies in its fringe tracking sensitivity. The challenge is therefore to maximize this sensitivity while minimizing the spreading of the signal on the detector. This is the goal at the core of the hierarchical fringe tracking (HFT) concept. We present the laboratory characterization of the $2^{nd}$ generation HFT chips operating in the near infrared (H band) for up to 4 telescopes with more linear phase and group delay estimators, allowing a strong simplification of the tracking algorithm. We show a comparison between theoretical intensity outputs for an optimized phase delay chip, and the laboratory measurements of two chips with different designs. The results do not reach the expectations but get close to them with the 10-outputs chip. Ultimately, this new chip is intended for implementation on the VLTI, CHARA or on the future Xuyi 100m-baseline Stellar Interferometer using three telescopes.
Read the full abstract
High Energy Physics (Phenomenology)
Exothermic Dark Matter at LZ
We investigate whether the high-energy nuclear recoil reported by the LUX-ZEPLIN (LZ) experiment can be explained by exothermic dark matter. We realize this framework in a minimal inelastic dark photon model and show that both freeze-out and low-reheating-temperature freeze-in cosmologies can reproduce the observed rate, with GeV-scale mediator masses and small kinetic mixing. We find that the LZ event can occur in a preferred parameter region with dark matter mass around $m_\chi \sim 30-200~\mathrm{GeV}$, and mass splitting $\delta \sim 0.5-1~\mathrm{MeV}$. The scenario is largely insensitive to halo uncertainties and has robust predictions for different targets. In particular, argon and germanium detectors are expected to observe events at energies above the current xenon window.
Read the full abstract
Algebraic Geometry
Semiorthogonal decompositions for families of twisted flag varieties
We show that the derived categories of smooth families of twisted generalized flag varieties admit semiorthogonal decompositions into derived categories of twisted sheaves over the base. In particular, we obtain a categorification of Panin's computation of the Quillen K-theory of twisted flag varieties. The main ingredient is a generalization of the Samokhin-van der Kallen semiorthogonal decomposition for representation categories of parabolic subgroups from split simply connected semisimple groups to arbitrary quasi-split semisimple groups, including non-simply connected cases.
Read the full abstract
Image and Video Processing
Real-World Multi-Modal and Longitudinal Lung Cancer Dataset
Multi-modal learning has demonstrated strong potential in medical applications by integrating heterogeneous data sources such as medical imaging, clinical records, and genomics to improve predictive performance and support clinical decision-making. However, advances in this area are often constrained by two key challenges: the limited availability of well-curated, ready-to-use datasets that accurately reflect real-world conditions, where medical data are frequently collected inconsistently and are often incomplete; and the inherent difficulty of integrating heterogeneous data modalities. In this work, we introduce a newly curated multi-center, multi-modal, and longitudinal dataset designed to support the evaluation of a wide range of learning pipelines under realistic conditions. The dataset comprises a total of 1,365 lung cancer patients and has three imaging modalities (whole-slide images, CT scans, and PET scans), structured clinical data, transcriptomic, and longitudinal follow-up and treatment information. For each imaging modality the dataset contains more than one instance. Moreover, the dataset exhibits substantial and non-uniform missingness across modalities, making it well-suited for studying robust multi-modal fusion strategies. We further provide both uni-modal and multi-modal benchmarks on the task of 12-month overall survival prediction, disease-specific survival, as well as longitudinal benchmark of hazard prediction under severe missing data. Our results show that, despite high levels of missingness, integrating complementary modalities consistently improves predictive performance over uni-modal approaches, highlighting the value of multi-modal fusion in realistic clinical settings. The dataset and benchmark code are available at https://github.com/ritacmendes/MMIST-LUNG.
Read the full abstract
Quantum Physics
Optimal inequalities for completely bounded polynomials and the limitations of quantum query algorithms
We consider the problem of establishing limitations on the power of quantum query algorithms via the completely bounded polynomial method. In particular, we prove several optimal functional inequalities involving different notions of completely bounded polynomials. These inequalities lead to limiting theorems for the power of quantum query algorithms that improve on prior works. 1. An optimal root-influence bound for block-multilinear polynomials. Prior work showed that block-multilinear polynomials $p$ of degree $t$ satisfy a root-influence bound, $\|p\|_{\text{cb}}\geq \sum_i \sqrt{\mathrm{Inf}_i[p]}/t^2$, which is stronger than the bound appearing in the Aaronson-Ambainis conjecture. We find the optimal constant in that inequality: $\|p\|_{\text{cb}}\geq \sum_i \sqrt{\mathrm{Inf}_i[p]}/t$. Since the amplitudes of quantum algorithms that query disjoint blocks of inputs-such as $t$-fold forrelation- are block-multilinear polynomials with $\|p\|_{\text{cb}}\leq 1,$ our inequality shows that they satisfy $t\geq \sum_i\sqrt{\mathrm{Inf}_i[p]}$. We prove that this inequality yields both a more efficient classical simulation than prior results based on the Aaronson-Ambainis argument, and a qualitative improvement: all classical queries are nonadaptive. 2. Optimal Fourier growth of the highest level of quantum query algorithms. We show that for every polynomial $p$ defined on $\{-1,1\}^n$ of degree $2t$, the Fourier Growth at the level $2t,$ namely $\|\widehat p_{2t}\|_{\ell_1}$, satisfies $\|\widehat p_{2t}\|_{\ell_1}\leq (en/(2t-1))^{\frac{2t-1}{2}}\|p\|_{\text{cb}}$. This is optimal up to the factor $e$, as witnessed by $2t$-fold forrelation. As quantum query algorithms that make $t$ queries (to the whole input) satisfy $\|p\|_{\text{cb}}\leq 1$, this yields a Fourier growth bound for these algorithms, partially resolving a question by Girish (STOC, 2026).
Read the full abstract
Geometric Topology
On the Mahler measure and root distribution of the $Q$-polynomial of links
We study the roots and the Mahler measure of the $Q$-polynomial of links. We first consider links obtained by adding twists to a pair of parallel strands. We prove that the Mahler measure of the transformed $Q$-polynomial converges as the number of twists increases. We also show that all but a uniformly bounded number of distinct roots of the $Q$-polynomial approach the real interval $[-2,2]$. This behavior is different from that of the roots of the Jones polynomial under twisting. Numerical experiments on prime knots lead us to a conjecture about the roots of the transformed $Q$-polynomial of alternating knots. Finally, we compare real and unit-circle roots of the Alexander, Jones, and $Q$-polynomials, and give an infinite family of $2$-bridge links whose $Q$-polynomials have only real nonzero roots.
Read the full abstract
Systems and Control
Risk-Aware Optimal Control with Rulebooks
We consider safety-critical control problems involving multiple requirements with different priorities and uncertainty in their evaluation. We represent these requirements using risk-aware rulebooks, where each requirement is assigned a risk measure and an acceptable threshold, and a priority relation is defined among the requirements. Each requirement induces a risk-evaluation function that maps a policy to the risk associated with its violation. We formulate risk-aware optimal control with rulebooks as a lexicographic optimization problem over excess risks and develop an anytime filtering and branch-and-bound algorithm that progressively tightens the certified optimality gap while characterizing the corresponding set of policies at each priority level. The algorithm returns a policy together with these gaps, which bound its suboptimality. We prove that these gaps are valid for any finite computational budget and, under additional assumptions, converge to zero as the computational budget increases. We evaluate the algorithm on a synthetic benchmark with a known optimum and a realistic highway-merging simulation with CVaR-based collision, rear-braking, headway, and comfort rules.
Read the full abstract
Artificial Intelligence
What Matters in On-Policy Distillation? A Perspective on Data Efficiency and Data Selection
On-Policy Distillation (OPD) has emerged as a widely adopted post-training paradigm for enhancing large language models in reasoning domains. However, the data-centric mechanisms in OPD remain relatively underexplored. This paper presents an empirical study of data efficiency and data selection in OPD. We begin by investigating an extreme setting: training OPD on only one example, namely 1-shot OPD. Surprisingly, we find that 1-shot OPD is consistently effective across all sampled training examples and harder examples often yield superior performance gain. We next investigate what actually drives the student model's improvement in the training data. Our analysis reveals that the improvement is not driven by high token entropy, but the longer CoT paths which hard problem naturally generate. Training on longer CoT can help maintain closer alignment with the teacher over a long reasoning horizon, and learn critical thinking patterns usually missing in short CoTs, such as reflection (e.g., "Alternatively"). Based on these insights, we propose a simple data selection method that selects only hard examples for training, where even "unsolvable" examples that completely exceed the teacher's capability can be successfully used. Our experiments conducted on five models ranging from 1.5B to 8B show that training the student model on only 8 selected hard examples matches the performance of the 17K dataset baseline. Code and models will be publicly available.
Read the full abstract
Complex Variables
Transverse Analytic Envelopes and the Baum–Bott Rationality Conjecture
We study holomorphic first integrals near singular leaves through transverse analytic envelopes. For diagonal actions these are normal toric quotients, and complete semiglobal integrability along singular orbits with independent support weights is equivalent to maximal resonance rank and finite monodromy. For commuting fields in convergent simultaneous Poincaré–Dulac normal form, induced derivations determine the first-integral algebra. Labelled quotients and logarithmic residues reconstruct weights at maximal rank; Grassmannians describe the ambiguity otherwise. Polynomial examples show losses invisible to prescribed finite jets. For the Baum–Bott rationality conjecture, we prove rationality and an explicit residue formula for smooth compact singular components with product charts and a common invertible semisimple transverse linear part of maximal resonance rank. Equivariant cohomology gives higher-degree vanishing for affine diagonal torus actions, including intersecting singular strata.
Read the full abstract
Accelerator Physics
Indirect Integration of Longitudinal and Transverse Wake Potentials for Unequal Beam Pipes and Arbitrary Beam Velocity
Indirect integration replaces the long uniform beam-pipe parts of a wakefield calculation by field problems in the pipe cross sections. Earlier ultrarelativistic methods were developed mainly for the longitudinal wake, whereas the transverse wake was usually obtained from the Panofsky–Wenzel theorem. We derive indirect formulas that complete a transverse Lorentz-force integral already accumulated in a time-domain calculation. At $\beta=1$, each semi-infinite tail is found from a Dirichlet Poisson problem driven by $E_z$ and a Neumann Poisson problem driven by $cB_z$, describing the TM and TE contributions, respectively. We obtain both a fixed-time moving-window representation and a fixed-plane time-history representation for equal or unequal input and output pipes. The method is then extended to a rigid bunch moving with constant velocity $0<\beta c<c$. The longitudinal correction satisfies an anisotropic elliptic equation in $(x,y,s)$ and provides an additional source for the transverse TM problem. In a two-port finite-reference convention, the complete fields, including space charge, are integrated directly between two fixed planes, while the semi-infinite tails are calculated after subtraction of the stationary field in each pipe. The resulting Panofsky–Wenzel relation contains the difference of the stationary transverse electric fields at the two ports. Numerical tests for an unequal rectangular step-out at $\beta=1$ and $\beta=0.8$ confirm the transverse indirect integration and the unequal-pipe boundary term.
Read the full abstract
Probability
Asymptotic Behaviour for Isotropic Pearson Random Walks
In the Pearson random walk, the direction of the i-th step is a random variable uniformly distributed on the unit sphere of R^d, and its length is a non-negative random variable. We prove that, when d goes to infinity, the Euclidean norm of the walk converges almost surely to the Euclidean norm of the vector of step lengths, whose joint law is arbitrary. Further, we investigate the same convergence result in a more general framework, where both the number and the lengths of the steps depend on the dimension. The results in the present paper can be applied to approximate the distribution of some classical Pearson random walks, for example, in the Dirichlet case.
Read the full abstract
Machine Learning
Phase Transition Frequency as a Training Time Predictor of Test Accuracy in ResNets
The number of discrete class-separability jumps observed during ResNet finetuning is examined empirically as a predictor of final test accuracy. Across 75 experiments spanning four benchmarks (CIFAR-10, CIFAR-100, TinyImageNet, and CIFAR-10-C) and three architectures (ResNet-18, ResNet-50, and ResNet-101), with five to ten seeds per configuration, a strong within-dataset negative correlation is obtained on standard i.i.d. classification benchmarks: \(r = -0.84\) on CIFAR-10 (\(p < 10^{-8}\), \(n = 30\)) and \(r = -0.87\) on CIFAR-100 (\(p < 10^{-5}\), \(n = 15\)). Under distributional stress, the relationship attenuates: TinyImageNet yields \(r = -0.45\), and the CIFAR-10-C corruption benchmark yields \(r = -0.19\). Two additional analyses discipline the empirical claim. A partial correlation controlling for architecture depth, treated as a linear covariate, shows that on CIFAR-100 the transition count retains statistically significant predictive power (\(r_{\mathrm{partial}} = -0.69\), \(p = 0.007\)); the corresponding result under the stricter categorical conditioning is not established at \(n = 15\). A comparison against six alternative training-curve signals shows that transition count achieved the strongest correlation among the evaluated signals on CIFAR-100 and one of the strongest on CIFAR-10, but is dominated by other signals on the two stressed benchmarks. The comparison is restricted to training-curve-level signals; comparisons against effective rank, Hessian sharpness, Fisher information, margin, and neural-collapse measures, which are the strongest competitors in the current literature, are not part of the present study and remain open. The observation is presented as an in-distribution training-quality probe among a family of candidate probes, and an inexpensive detection procedure suitable for logging alongside a standard training loop is provided.
Read the full abstract
Analysis of PDEs
Global well-posedness of the 3D Navier-Stokes equations with helical $L^1$ vorticity
We show that the Navier-Stokes equations on $\mathbb R^2\times\mathbb T$ are globally well posed for initial vorticities that are both helically symmetric and integrable. This class of data typically generates velocity fields of infinite kinetic energy, so this result is not covered by the classical finite-energy well-posedness theory. This result is supercritical from the perspective of the three-dimensional Navier-Stokes scaling, and it is made possible by the special structure of helical flows using time-weighted Kato-type spaces.
Read the full abstract
Superconductivity
Dynamical Reduction of Two Series Josephson Junctions to a Synthetic High-Transparency Josephson Element
Two conventional Josephson junctions connected in series can reproduce, in the static limit in which the currents through the capacitive and resistive channels are negligible, the current-phase relation of a single effective weak link with tunable transparency. Therefore, the two-junction series can be treated as a single synthetic high-transparency element. Here, we investigate to what extent this mapping remains valid under finite-frequency drive and retaining the junctions' resistive and capacitive terms. The full resistively and capacitively shunted junction equations are compared with an effective synthetic element with tunable transparency that retains the synthetic tunable-transparency current-phase relation together with effective capacitive and dissipative terms, thus reducing the two second order degree of freedom system to a single second order degree of freedom. The resulting single-element dynamics is compared with the complete two-junction system under ac excitation. The agreement is quantified through a normalized root-mean-square error between the full and effective voltage waveforms. A broad low-error region is found at low drive frequency, while pronounced deviations emerge as the drive frequency approaches the relevant plasma-frequency scale and at larger drive amplitudes. The results provide a quantitative dynamical criterion for using the reduced single-element description of a synthetic high-transparency Josephson element in superconducting circuits.
Read the full abstract
Game Theory
Cutting Down the Tower: Single-Exponential Envy-Free Cake Cutting
Envy-free cake cutting is a central problem in fair division with a striking divide between existence and computation. Classical topology guarantees that envy-free allocations exist, yet finding one efficiently (in the Robertson-Webb query model) has resisted decades of work. The seminal work by Aziz and Mackenzie established the existence of a bounded protocol for every $n$, but its query bound is $n^{n^{n^{n^{n^n}}}}$. Subsequently, Sokolov modified their protocol and sharpened its analysis, obtaining a bound of $n^{8n^2(1+o(1))}$ queries, the best deterministic upper bound known before this work. In contrast, the general lower bound, due to Procaccia, is merely $\Omega(n^2)$. We close much of this gap with a deterministic protocol using $n^{O(1)}2^n$ queries, the first single-exponential bound for general valuations. Like Aziz and Mackenzie, we repeatedly allocate cake without envy until some agents can leave with their pieces, guaranteed not to envy anyone who stays, even if a staying agent receives all the remaining cake. To reach this state, Aziz and Mackenzie generate $n^{n^{n^n}}$ partial allocations so that enough of them can be redistributed without conflict. Instead, we show that only polynomially many partial allocations suffice. Our key idea is to “prune” scarce allocations upfront: if only a few allocations can help a certain pair of agents, we discard them and handle that pair another way, leaving plenty of allocations for every remaining pair so each gets a distinct one. As a result, we make only polynomially many calls to Aziz and Mackenzie's subroutines, and this gives us our $n^{O(1)}2^n$ bound.
Read the full abstract
Artificial Intelligence
The Mirror Agent Model: a Bayesian Architecture for Interpretable Agent Behavior
In this paper we illustrate a novel architecture generating interpretable behavior and explanations. We refer to this architecture as the Mirror Agent Model because it defines the observer model, that is the target of explicit and implicit communications, as a mirror of the agent's. With the goal of providing a general understanding of this work, we firstly show prior relevant results addressing the informative communication of agents intentions and the production of legible behavior. In the second part of the paper we furnish the architecture with novel capabilities for explanations through off-the-shelf saliency methods, followed by preliminary qualitative results.
Read the full abstract
Computation and Language
Can Large Language Models Anticipate Behavioral Responses to Social Policies? A Case of Pension Enrollment Prediction among China's Flexible Workers
Assessing the impacts of social policy changes is a widely acknowledged challenge for policymakers. Econometric methods can be unreliable when extrapolating to hypothetical scenarios, while field pilot programs are highly costly. In this paper, we propose using large language models (LLMs) as policy-assessment tools adapted from general-purpose models. We present FlexPension-LLM, the first domain-specialized large language model for a hierarchical pension-enrollment prediction task among flexible workers in China, and introduce DKI-RDistill, which injects policy-grounded cues into the prompt, including Probit-derived marginal effects and hukou-province pension rules. The method then uses LoRA/SFT to distill rationale-augmented supervision into an open-weight MoE student, with teacher errors corrected by regenerating those cases under ground-truth labels. On a CHFS 2019 blind split, FlexPension-LLM achieves 0.9316 Composite F1, surpassing its Claude Sonnet 4.5 teacher and 15 of 17 baselines, and is statistically indistinguishable from Claude Opus 4.6. Across four external surveys, it averages 0.7549 Composite F1 and shows the narrowest performance range among the strongest systems. Component analysis shows that gains come mainly from policy-grounded cue injection and error-filtered supervision, while rationales provide decision traces that can be checked against policy rules.
Read the full abstract
Mesoscale and Nanoscale Physics
Electrostatic splitting of an Edge Magnetoplasmon Resonator
Edge-magnetoplasmon resonators have been proposed as a powerful tool to detect anyons by introducing a quantum point contact into an isolated quantum Hall system probed via radiofrequency radiation. In this paper, we study the effect of a quantum point contact embedded within an edge-magnetoplasmon resonator and how its polarization influences the propagating magnetoplasmonic mode. Combining dc and rf measurements, we unambiguously evidence the signature of both integer ($\nu = 1$ and $2$) and fractional quantum Hall states ($\nu = 4/3$ and $2/3$) within the radiofrequency transmission signal. Using electrostatic gating, we determine the physical parameters characterizing the electrostatic edge of an AlGaAs/GaAs based two-dimensional electron gas. We extract the dependence of the cavity perimeter with the gate voltage of the quantum point contact and fully characterize the path followed by edge magnetoplasmons in this system. Finally, we provide a geometric model in good agreement with experimental results.
Read the full abstract
Geometric Topology
Two-numbers and Euler characteristics for quandles
Quandles can be regarded as generalizations of symmetric spaces. In the theory of symmetric spaces developed by Chen and Nagano, there is an interesting relationship between the two-number and the Euler characteristic. The two-number is a Riemannian geometric invariant that can also be characterized in terms of point symmetries, whereas the Euler characteristic is a topological invariant. The aim of this paper is to initiate the study of quandle analogues of Chen–Nagano theory. In particular, we investigate relationships between the two-numbers and the Euler characteristics of quandles, and provide examples of finite quandles that either satisfy or fail to satisfy properties analogous to those of symmetric spaces. These examples are constructed from directed simple graphs labeled by abelian groups.
Read the full abstract
Software Engineering
An Empirical Study on Learning Paths and Gender Dynamics in Scrum Master Roles
Context: Agile development methodology has been widely adopted by industry and the demand for experienced professionals in Agile-related roles is persistently high. Objectives: We focus on the learning path for a Scrum Master role in multicultural software companies and investigate the role in relation to team size, together with the learning process for a career path, and how companies monitor soft skills development. Method: We conducted our study in two phases, two qualitative surveys (interview studies) and performed a qualitative and quantitative data analysis of the results. Conclusions: Our results identified that the need for a Scrum Master (SM) depends on the size of the team, with our study indicating a six-member limit. There is no overall standardized process for soft skills learning or metrics to measure progress. Some companies measure soft skills based on feedback received from the client or from the team, and other companies are taking both types of feedback into consideration. Many learning initiatives, especially on soft skills for an SM role, were based on actions of the employees. The version of Record of this contribution is published in Software Engineering and Advanced Applications. SEAA 2025. Lecture Notes in Computer Science, vol 16083. Springer, Cham. and available online at: check DOI.
Read the full abstract
Superconductivity
Ni-O hybridization as a stabilizer for $s^{\pm}$ superconductivity in La$_3$Ni$_2$O$_7$: a DFT+RPA study
The superconducting gap symmetry of high-pressure bilayer nickelates remains under debate, with weak- and strong-coupling approaches yielding different pairing tendencies. In this work, we investigate how the weak-coupling treatment of electronic states away from the Fermi level influences magnetic fluctuations and superconductivity in La$_3$Ni$_2$O$_7$. We employ a full-spectrum model based on orthonormalized projections of Kohn-Sham states onto local Ni-$e_g$ orbitals, which preserves the density-functional band structure while redistributing spectral weight over a wide energy range. Compared to a low-energy description, this approach yields enhanced interlayer spin fluctuations and a commensurate magnetic instability. Within a spin-fluctuation framework, these features favor a sign-changing $s^\pm$ superconducting state, whereas low-energy models tend to stabilize $d$-wave pairing. Our results suggest that interlayer coupling in full-energy models may play an important role in shaping the predicted pairing symmetry of bilayer nickelates.