Artificial Intelligence
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Sunday 8 November
Large language models (LLMs) perform well on general coding tasks, but their capabilities in application-layer embedded software development remain underexplored. Existing embedded benchmarks primarily focus on hardware-in-the-loop validation in simulated environments. We introduce RealEmbedBench, a benchmark of 5,142 samples covering three application-layer tasks: code generation and completion, state machine modeling and implementation, and cross-OS code translation. We evaluate six representative LLMs and find substantial performance differences across tasks and models. Project context, especially the local code window, significantly affects code generation and completion. Models perform reasonably well at state machine modeling but struggle to translate these models into executable code. Cross-OS translation performance varies substantially by direction. Translation from Harmony to RT-Thread is particularly challenging, with an average pass rate of 0.1667. These findings suggest that developers should consider local code context when using LLMs for code generation and completion and select models according to the translation direction. They also motivate the use of more real-world state machine implementation data in LLM training to narrow the gap between modeling and executable code.
Quantum Physics
Quantum State Reconstruction from Sklar-Type Relational Data
Sklar's theorem in classical probability theory rigorously separates the marginal statistics of a joint probability distribution from its dependence structure data. We apply this perspective to the context-dependent classical probability tables induced by the Born rule for a quantum state. For each Born probability table of a given quantum state, we discard the marginal statistics of the individual parties, retaining only the relational data among them. Yet we prove that, when these classical relational data are collected over all contexts, the nontrivial quantum compatibility constraints inherent in quantum mechanics suffice to reconstruct any irreducible multiqubit quantum state itself. This means that the local information of the subsystems, discarded table by table, is fully recovered. We call this quantum-state reconstruction from Sklar-type relational data.
Read the full abstract
Databases
PLDP-IDS: Personalized Local Differential Privacy for Infinite Data Stream
Existing $w$-event local differential privacy (LDP) mechanisms for streaming data impose a uniform privacy window and budget, forcing all users to adopt the most stringent privacy requirement. We study personalized continual frequency estimation under LDP, where each user specifies her own window size and privacy budget, and identify two challenges. First, a single sampling size cannot balance sampling and perturbation errors for privacy groups with heterogeneous requirements. Second, borrowing estimates from groups with larger budgets reduces variance but introduces bias when privacy preferences correlate with the data. We first propose Optimal Population Selection (OPS) and two OPS-guided mechanisms, PLPD and PLPA, which use one privacy-compatible sampling size for all users. Since one size cannot fit all groups and pooling their estimates is biased under privacy–data correlation, we further develop PLPD$^+$ and PLPA$^+$, which combine group-aware population scheduling with bias-aware cross-group fusion while preserving personalized $w$-event LDP. Experiments on real-world and synthetic streams show that, compared with the direct personalized extension of LDP-IDS, PLPA$^+$ reduces the average mean squared error (AMSE) by 52.3% and 40.8% on average, respectively, and remains 4.3–5.1$\times$ more accurate than PLPA under strong privacy–data correlation.
Read the full abstract
Numerical Analysis
A DG method for the pseudostress-velocity formulation of the Stokes-Brinkman eigenvalue problem
We introduce a discontinuous Galerkin method for the approximation of the eigenvalues and eigenfunctions of the Stokes-Brinkman eigenvalue problem. The method is based on a mixed formulation in which the pseudostress tensor and the fluid velocity are the main unknowns, while the pressure can be recovered through a post-processing procedure applied to the pseudostress tensor. Using discontinuous polynomial spaces of degree $k\geq 1$, we propose a consistent symmetric interior penalty discretization for the approximation of the spectrum. Relying on the spectral theory for non-compact operators, we establish convergence of the discrete spectral approximation and derive error estimates for the method. A series of numerical experiments is presented to assess its computational performance. On the one hand, we investigate the convergence rates for different polynomial degrees, computational domains, and values of the permeability. Furthermore, we study the influence of the stabilization parameter on the discrete spectrum by comparing different choices of this parameter and examining their impact on the occurrence of spurious eigenvalues.
Read the full abstract
Sound
When Normality Shifts: Multi-Level WavLM Modeling for Open-World Audio Deepfake Detection
Audio deepfake detectors trained with known synthesis attacks can learn generator-specific shortcuts that fail as synthesis systems and acoustic conditions evolve. We investigate an attack-free alternative: a normality model trained exclusively on bona fide speech, which scores synthetic audio by its deviation from authentic representations. We propose MLN-WavLM, a method that applies shrinkage-Mahalanobis modeling to pooled hidden states from a frozen large-scale, pre-trained speech processing WavLM Base+ architecture. Evaluated on ASVspoof 2019 LA, our best attack-free variant achieves an 8.45% Equal Error Rate (EER), closely tracking a fully supervised WavLM-HM-Conformer baseline (7.86%). However, when subjected to the ASVspoof 2021 DF domain shift, performance drops to 22.12% EER, exposing vulnerabilities in the bona-fide-calibrated operating point. Layer-wise analysis reveals that early WavLM representations dominate the ensemble. To test real-world viability, we simulated deployment streams using unlabeled threshold adaptation. While robust median/MAD recalibration easily handles in-domain contamination up to 10%, recalibrating under the 2021 DF domain shift triggers a total collapse of low-FPR recall. We conclude that while frozen source-domain normality can effectively rank deepfakes, it does not yield a stable open-world decision rule, and downstream threshold adaptation cannot rescue upstream ranking degradation.
Read the full abstract
Robotics
Off-Policy RL with Online Inference Tuning: My Solution to the CoRL 2026 RoboEval Challenge
The CoRL 2026 RoboEval challenge has eight bimanual tasks, and it ranks a policy not only on whether it succeeds but on how: every stage of the task, done safely, efficiently and with both arms in coordination, scored against human demonstrations. My solution has two parts. The generic part, carried over from my LeHome challenge solution, is off-policy RL on a $\pi_{0.5}$-based policy that is also its own value function. It is trained with AWR and RECAP on advantages of the exact official score, in a loop where every rollout is at the same time a Thompson-sampling trial of about 20 inference parameters. Out of the box, it took the policy to 0.85 RDS on held-out seeds in two days (and it was still very far from convergence). The competition-specific part adds data that local exploration cannot find: generated demonstrations that reach the full reward, and recovery episodes that start from the policy's own failures. The submitted policy reaches 0.990 RDS on held-out seeds and 0.987 on the official seeds.
Read the full abstract
Algebraic Topology
On Poincaré spaces which are not manifolds
This expository note gives a short derivation of the classical Spivak normal fibration formula for the total space of a spherical fibration over a Poincaré duality space. We review the associated smooth, PL, and topological manifold realization criteria and illustrate them with examples from the stable stems. The principal application combines these criteria with the completed solution of the Kervaire invariant problem to identify the exceptional stems in a specified family of three cell Poincaré spaces.
Read the full abstract
Systems and Control
Local Certificate Repair for Asynchronous Controller Updates in Recommender Systems
We study how to adapt an existing distributed recommendation controller when local dynamics or objective weights change. Starting from a quadratic stability and cost certificate, we construct positive semidefinite corrections with finite spatial support and an explicit residual. A joint admission test allocates the available dissipation margin to these residuals and to interactions among changes and corrections, including pending updates. The test covers complete declared parameter ranges, allowing different operators to progress independently. For a fixed compatible batch, the corrections define a common storage that guarantees exponential stability and bounds the cost accumulated throughout deployment. When the admitted collection evolves, a reservation protocol preserves the certificate and a rate bound on storage increases ensures exponential convergence. Regional linear-quadratic regulator (LQR) design proposes new gain rows for the operators' own users, and a separate residual-based calculation can certify that the final gain lowers the expected cost. A recommendation-network example illustrates the tradeoff between repair depth, data access, and the benefits of deploying an update.
Read the full abstract
Information Theory
Sharp Rates for Optimal Recovery of Bandlimited Functions from Localized Oversampling
It is well-known that one can recover a function $f$ bandlimited to $[-\delta,\delta]$ ($0<\delta<\pi$) from its finite oversampling data $f(j),-n+1\le j\le n$, with an approximation error exponentially decaying to zero as $n$ tends to infinity. Various regularized Shannon sampling formulas have been proposed to achieve such exponential approximation. A recent sinh regularized formula achieved so far the best rate $$ O(\frac1{\sqrt{n}}\exp(-(\pi-\delta)n)). $$ We establish that the optimal recovery algorithm in the worst-case scenario achieves an approximation error $\rho_{n,\delta}$ satisfying $$ \frac{c_*}{\sqrt{n}}(\tan\frac\delta 4)^{2n+1/2}\le \rho_{n,\delta}\le \frac{C_*}{\sqrt{n}}(\tan\frac\delta 4)^{2n+1/2}, 0<\delta\leq\pi, n\geq1, $$ where $c_*,C_*$ are explicit positive constants independent of $n$ and $\delta$. This result identifies the best possible exponential rate to reconstruct bandlimited functions from two-sided finite oversampling data. In other words, $$ \rho_{n,\delta} \asymp \frac{\sqrt{\tan\frac\delta 4}}{\sqrt{n}}\exp(-c_\delta(\pi-\delta)n), $$ where $c_\delta$ is strictly decreasing on $(0,\pi]$ and $$ \lim_{\delta\to\pi^-}c_\delta=1,\ \lim_{\delta\to 0^+}c_\delta=+\infty. $$ For one-sided sample data $f(j)$, $-n+1\le j\le 0$, we also obtain the sharp approximation order of optimally recovering function values on $[0,1]$: $$ \rho_{n,\delta}^-\asymp \sqrt{\delta/\pi}\sin^n(\delta/2),\ 0<\delta<\pi,\ n\ge1, $$ which improves the celebrated result of H. J. Landau on extrapolation of bandlimited functions.
Read the full abstract
Machine Learning (Statistics)
Identifiable alignment of unpaired representations
Aligning two models without paired data requires a distance that both share, yet existing methods use Euclidean or spherical geometry. This is either an arbitrary choice or imposed by their loss function. We show that, under a precise form of the manifold hypothesis, the pullback of the data metric gives such a distance for any two models of one data-generating process. This holds whether the models are encoders, decoders or both, and whatever their architecture, loss, prior or training data. The alignment map preserves this distance and the distribution of representations, and is identifiable if and only if the data distribution breaks every symmetry of the data manifold. We show that this symmetry breaking holds generically. Identifiability makes the alignment coupling the unique optimum of Gromov–Wasserstein and of its weighted combinations with a linear program. The linear program alone has the same optimum if almost every data point has a distinct distribution of distances. The coupling, and with it the alignment map, is recovered exactly from permuted samples, even without identifiability, and consistently from independent ones. Our method, Neural Gromov–Monge Alignment (NGMA), uses no pairs or labels and outperforms every unpaired and label-informed baseline. It relies on pullback distances to align image models with different priors and single-cell models trained on disjoint cells, where Euclidean distances fail.
Read the full abstract
Optimization and Control
From Imperfect Information to Controlled Steps: Stochastic Optimization under Heavy-Tailed Noise and Compression
Stochastic optimization relies on information whose statistical quality, availability, and computational cost can vary substantially. Heavy-tailed observations challenge concentration, compression distorts communicated vectors, and generalized smoothness changes how far an algorithm can move. This dissertation studies these difficulties through two linked design decisions: how to construct an informative search vector and how to convert it into a controlled step. We first develop high-probability methods under bounded central $p$-th moments, $p\in(1,2]$, allowing infinite variance when $p<2$. Clipped gradient, accelerated, extragradient, and gradient descent-ascent methods come with guarantees for stochastic minimization and variational inequalities, with smoothness and moment conditions needed only where the iterates lie with high probability. For composite and distributed problems, shifted and innovation clipping preserve the fixed point that direct clipping can disturb, while retaining linear speedup. We then characterize the value of richer stochastic information: lower complexity bounds quantify when same-sample gradient differences and stochastic Hessian information improve heavy-tailed nonconvex optimization, and normalized momentum variance reduction and Hessian-corrected momentum attain matching or near-matching in-expectation bounds. Separately clipping gradients and their correction terms yields batch-free high-probability methods. For compressed distributed optimization, we prove convergence of normalized error-feedback methods under generalized smoothness without a uniform gradient-dissimilarity assumption, and develop parameter-agnostic variants with five momentum estimators. Finally, we extend the estimator hierarchy to linear minimization oracles over arbitrary norm balls, recovering the Euclidean rates under estimator-specific generalized smoothness.
Read the full abstract
High Energy Physics (Phenomenology)
After the Higgs: Looking for the Universe Behind the Particle
Fourteen years after its discovery, the Higgs boson has changed roles. It was once the object of the search, but today it is an instrument with which we interrogate Nature's vacuum, test quantum field theory and look for new principles. This shift comes at a time when, unlike before the Higgs discovery, we have no theoretical argument guaranteeing new physics within experimental reach. At the same time, securing long-term support for fundamental science is becoming increasingly challenging. I revisit the path from a theoretically motivated discovery to a precision programme, explain why the Higgs is such an unusual quantum field, and argue that this same peculiarity makes it a gateway to cosmology and other sectors. I then turn to a less technical question: how should a field plan for exploration when no theorem guarantees the next discovery? My answer is to pursue precision without pretending to know in advance which measurement will matter, sustain a diverse experimental portfolio, protect theoretical risk-taking, and make the public case for fundamental exploration on its own terms.
Read the full abstract
Machine Learning
MOMENT-DP: Momentum Aware Evaluation of Correlated Noise for Differentially Private Training
Temporally correlated noise can substantially improve the utility of differentially private optimization, yet candidate mechanisms are commonly designed or evaluated using only partial information about the optimization process, such as workload structure, learning-rate schedules, or local curvature. We study whether persistent optimizer state provides additional information that can change the relative utility of privacy-valid correlated-noise mechanisms. We introduce MOMENT-DP, an optimizer-state-aware framework that models how temporally correlated privacy noise propagates through momentum-based optimization. For locally quadratic objectives, MOMENT-DP expresses privacy-induced final loss through an optimizer transfer operator that jointly captures curvature, learning-rate scheduling, momentum, and temporal noise covariance. NoiseCurve's curvature-aware quadratic criterion is recovered exactly as the zero-momentum, constant-learning-rate special case. Changing only momentum while holding curvature, schedule, privacy calibration, and candidate mechanisms fixed induces genuine mechanism-ranking reversals. The effect persists outside exact quadratic dynamics in per-example-clipped private logistic regression, while remaining unresolved in the neural regime we study. These results establish optimizer state as an under-modeled, decision-relevant axis in evaluating temporally correlated privacy noise.
Read the full abstract
Statistical Methodology
Model-assisted semi-supervised estimation of the Mann-Whitney probability under right censoring
We study semi-supervised estimation of the full-lifetime Mann-Whitney probability under right censoring using finite covariate-only samples from two populations. A model-assisted estimator is constructed by adjusting censored pair comparisons with differences between labeled and auxiliary feature means. Resource weights minimize first-order variance within the working projection class. Under joint independent censoring and regularity conditions, we establish consistency and asymptotic normality through a four-pool influence representation that accounts for censoring distributions estimated from the same labeled records. Consistent variance estimation from all four sample pools yields Wald confidence intervals. The efficiency analysis relates variance reduction to projected comparison signals and auxiliary sample sizes and characterizes the gap to the semiparametric bound. A fixed-horizon extension accommodates half-weighted ties and the terminal mass induced by capping follow-up. Simulations show increasing precision gains with auxiliary sample size under moderate or strong projected signals, with finite-sample losses in some settings with zero or weak projected signals. A five-year Rotterdam breast-cancer comparison illustrates precision gains and variability across label allocations when auxiliary samples are formed by controlled outcome withholding.
Read the full abstract
Cryptography and Security
Coding Agents Propagate Vulnerabilities in Code Bases
LLM-based coding agents are widely used to develop critical production software. Even though coding agents are known to occasionally write vulnerable code, there has not been any investigation into how code base vulnerabilities evolve over time under coding agent contributions. In this work, we close this gap by investigating whether and how vulnerabilities in existing code lead coding agents to introduce similar vulnerabilities in the code they write. We first evaluate coding agents on 96 tasks that require implementing a new backend application in a repository where other applications already exist. Second, we investigate repository-level code generation on 20 pairs of tasks in real C/C++ projects. In each task, the agent is asked to complete code critical to memory safety. We find that in both settings, insecure existing code leads to agents writing vulnerable code more often. In backend generation, insecure neighbors reduce the share of correct and secure implementations by 3.4% on average, and in C/C++ repository-level coding, the average gap between secure and insecure contexts reaches 35.4%. Prompting to follow existing code-style or to be careful about security increases, respectively decreases, how strongly agents propagate vulnerabilities, but neither stops them. When asked to explain their implementation choices, coding agents attribute them to following another code site in 76.5% of cases, even though they recognize the introduced vulnerability in 69.9% of cases. These findings show that agents use existing vulnerabilities as templates for new ones indiscriminatorily, turning each contribution into an opportunity to introduce more vulnerabilities. As such, as agents write a growing share of production code under decreasing human review, code bases can increasingly degrade in security if security vulnerabilities are not thoroughly fixed continuously.
Read the full abstract
Nuclear Theory
Anomalous charge conjugation and parity in (2+1)-dimensional relativistic hydrodynamics
We present a hydrostatic and first-order constitutive description of a mixed anomaly that involves $U(1)_Q$, $U(1)_F$, charge conjugation symmetry (acting only on $U(1)_Q$), and parity symmetry in $2+1$ dimensions. It is represented by a mutual Chern-Simons level $k_{QF}=\frac12 \,(\mathrm{mod}\ \mathbb Z)$. Adopting the already established multi-charge anomaly sector, we explicitly provide the information that this response is $odd$ under $\mathsf{C}$ and $\mathsf{P}$ symmetries separately, but $even$ under combined $\mathsf{CP}$ symmetry. We also provide a global inflow representative and compute the equilibrium cross-charge response which includes $\partial n_Q/\partial B_F=\partial n_F/\partial B_Q=1/4\pi \bmod 1/2\pi$. Using an analytic neutral-fluid example we show that this response is non-vanishing whereas the source-free hydrodynamic poles are invariant.
Read the full abstract
Strongly Correlated Electrons
Momentum-dependent phonon effects on spin-fluctuation $d$-wave pairing in bulk La$_6$Ni$_5$O$_{12}$
We study the effect of electron-phonon coupling on spin-fluctuation-mediated pairing in bulk La$_6$Ni$_5$O$_{12}$. A five-layer Hubbard model treated within the random-phase approximation yields a leading $d_{x^2-y^2}$ pairing state. Two phenomenological couplings motivated by oxygen breathing modes favor $s$-wave pairing when considered alone, while their attractive scattering between opposite-sign regions of the spin-fluctuation-induced gap suppresses $d$-wave pairing. Electronic charge screening strongly reduces this suppression. Comparing the $s$- and $d$-wave pairing strengths with and without screening shows that screening shifts their boundary in favor of $d$-wave pairing. Momentum-resolved projections locate the reduction of the unfavorable scattering processes. At the representative electronic interaction, $d$-wave pairing remains dominant.
Read the full abstract
Computer Vision
Semantically Aligned KV Cache Compression for Efficient Agentic Autoregressive Video Generation
Agentic autoregressive video generation combines semantic planning, sequential generation, and visual verification for controllable synthesis. Growing visual histories enlarge the key-value (KV) cache, increasing memory use and attention costs across candidate rollouts. Existing compression methods primarily optimize individual generators without the planner's semantic constraints, risking low priority for states supporting requested attributes or actions. We introduce SemanticKV, a training-free method for semantically aligned KV cache compression. To our knowledge, SemanticKV is the first KV cache compression method for agentic autoregressive video generation that explicitly uses planner-extracted semantic constraints to guide retention. Text cross-attention identifies constraint-related video positions, whose self-attention queries are collected into separate groups. Attention weights are averaged within each group and maximized across groups, giving each cached entry a high score when strongly used by any represented constraint. A single top-k selection retains high-scoring entries under the memory budget, leaving their keys, values, and positions unchanged. For fixed query groups and equal entry costs, this selection minimizes an additive upper bound on worst-group discarded attention. The method operates independently within each candidate without retraining or additional video generations. We evaluate four AR video backbones on three benchmarks for 5-, 30-, and 60-second generation. On Causal Forcing at a compression ratio of 0.7, SemanticKV obtains a VBench Total Score of 84.10, compared with 84.05 for the strongest compressed baseline and 83.99 for Full KV. It achieves a 1.22x speedup over Full KV and reduces KV-cache memory from 6.04 GB to 1.81 GB, a 3.35x reduction.
Read the full abstract
Group Theory
Finite soluble groups with Sylow subgroups of class at most two are IYB-groups
We prove that every finite soluble group whose Sylow subgroups have nilpotency class at most two is an involutive Yang–Baxter group ($\IYB$-group). This answers a question of Cedó and Okniński \cite{CedoOkninski2025}. The main method of the $\IYB$-structures has been reflected in \cite{MengBallesterJimenez2026}.
Read the full abstract
Machine Learning
How Much Training-Inference Mismatch Can On-Policy Distillation Tolerate?
On-policy distillation (OPD) trains a student language model on its own responses, guided by a fixed teacher. Generating these responses is expensive, so systems use quantized inference and asynchronous rollouts to reduce the cost. Both, however, introduce training-inference mismatch between the rollout engine and the model being trained. Such mismatch can destabilize reinforcement learning (RL), but its effects on OPD remain unclear. This paper investigates how much mismatch OPD tolerates, where its effects arise, and how training systems can use this tolerance. Measuring all mismatch sources on one scale, we examine quantization across three student-teacher pairs, with additional tests of weight staleness and injected noise on one pair. OPD tolerates substantial mismatch without meaningful loss of training quality but degrades at higher levels, revealing an empirical mismatch budget. Both the quality cost of mismatch and the mismatch introduced by quantization depend on the model, with some layers particularly sensitive to reduced precision. In controlled comparisons, OPD remains stable under rollout configurations that destabilize RL methods. Under weight noise, mismatch also shifts the prefixes reached during generation, introducing errors that per-token corrections cannot address. To put these findings into practice, we introduce a lightweight admission check that screens rollout configurations against a calibrated budget without requiring a full training run for each candidate. Model-specific FP8 recipes preserve training quality while providing capacity for at least $1.93\times$ as many KV-cache tokens as BF16. Combining FP8 with asynchronous rollouts reaches the same training quality $1.62\times$ sooner than synchronous BF16 training. Code is available at https://github.com/benyucong/train-inference-mismatch-opd
Read the full abstract
General Relativity and Quantum Cosmology
Pushing the accuracy of the quasi-circular baseline for the SEOBNR family of models
Gravitational-wave observations have opened a new era of astronomy, with accurate modeling of binary black hole signals becoming increasingly important for source characterization and astrophysical inference as detector sensitivities improve. We present SEOBNRv5iHM, an updated quasi-circular, aligned-spin inspiral-merger-ringdown gravitational-waveform model with higher-order modes, constructed within the effective-one-body formalism and building upon SEOBNRv5HM. The improvements mainly concern the late inspiral and the merger-ringdown: we remove the non-quasi-circular corrections used in previous models and instead introduce a phenomenological intermediate region between the inspiral and the ringdown, which allows the ringdown attachment time to be chosen independently for each mode, including at the peak of the mode's amplitude. The model is calibrated and validated against 591 quasi-circular aligned-spin waveforms from the third Simulating eXtreme Spacetimes catalog. For the $(2,2)$ mode, when considering the maximum unfaithfulness against numerical relativity over a total mass range of 10-300$M_\odot$, SEOBNRv5iHM achieves a median of $4.25\times10^{-5}$, a factor of 4.5 below SEOBNRv5HM, and a factor of 3-8 below other semi-analytical models, with $84\%$ of cases below $10^{-4}$. For $M\ge150 M_\odot$, the median drops to $9.7\times10^{-6}$, comparable to the numerical-relativity surrogate NRHybSur3dq8. For polarization mismatches at inclination $\pi/3$, with modes up to $\ell_\mathrm{max} = 5$, the median maximum unfaithfulness is $4.2\times10^{-4}$, a factor of 2.6 below SEOBNRv5HM, 1.9 below IMRPhenomTHM, and 6-11 below the other semi-analytical models. Since spin-precessing and eccentric models are built on top of the quasi-circular, aligned-spin baseline, these improvements will make future SEOBNR models more accurate, reducing waveform systematics in upcoming analyses.
Read the full abstract
Statistical Methodology
ZIO: Order-Based Structure Learning for Zero-Inflated Count Data under Data Heterogeneity
We propose a structure learning method for directed acyclic graph (DAG) models of zero-inflated multivariate count data. Although causal structure learning with zero-inflated Poisson models is identifiable, existing algorithms rely on local search over the DAG space, leading to poor scalability and a tendency to converge to suboptimal solutions. To address these limitations, we develop a novel order-based scoring framework that searches over variable orderings rather than directly over DAGs. Building on this framework, we further extend the method to multiple heterogeneous datasets by identifying a shared variable ordering while allowing each dataset to have its own edge set. We implement a stochastic hill climbing algorithm with a random-to-random (R2R) proposal operator to efficiently explore the ordering space. Simulation studies show competitive performance against existing methods and improved graph estimation accuracy as more heterogeneous datasets are jointly analyzed under a shared causal ordering. We apply the method to single-nucleus RNA sequencing data from a major depressive disorder study.
Read the full abstract
Machine Learning
Learning to Rank Propagators in Constraint Programming
The order in which propagators are executed can substantially affect the performance of constraint programming solvers, yet propagator priorities are typically determined by fixed, manually designed heuristics. In this work, we explore learning instance-dependent propagator priorities from static problem features. We formulate the task as a learning-to-rank problem, addressing the challenge that the quality of a priority assignment can be measured through solver runtime, while the desired priorities of individual propagators are not directly observable. A self-attention-based model jointly scores the propagators present in an instance and is trained using pairwise preferences derived from solver runtimes measured under randomized propagator priority assignments, with a Plackett-Luce-based ranking objective. Predicted rankings are mapped to the discrete priority levels of the Choco solver before search begins, without modifying its propagation or search procedures. We evaluate the approach on XCSP3 instances and demonstrate that propagator ordering can have a substantial impact on solving time. On a test set of 698 instances, picked out as most difficult by problem class, the learned priorities increase the number of solved instances from 634 to 654, reduce median solver runtime by 27.93%, and reduce aggregate penalized runtime (PAR1) by 12.65% compared with Choco's default priorities. We further observe that improvements vary across problem classes and show that a simple policy selecting between learned and default priorities further increases the number of solved instances. These results demonstrate the potential of learning instance-dependent propagator priorities to improve constraint programming performance, particularly on difficult instances.
Read the full abstract
Quantum Physics
One Bit, Many Records: Information Replication in Quantum Many-Body Dynamics
Information replication occurs when a single locally encoded classical bit becomes simultaneously readable across extensively many disjoint regions of a quantum many-body system. Here we identify dynamical selection as a mechanism for generating these records, whereby an unperturbed nonthermal reference remains locally recognizable while its locally perturbed counterpart spreads and relaxes under the same Hamiltonian. We establish an entropy-capacity criterion showing that a uniform area-law reference paired with a volume-law perturbed branch guarantees extensive records at a fixed, nonzero information threshold. An embedded-dimer chain demonstrates both record proliferation and increasing readout reliability with fragment size, while stabilized PXP dynamics extends the mechanism to a coherently recurrent reference. These results characterize information spreading through the spatial multiplicity of accessible records and provide experimental signatures measurable via fixed local readouts.
Read the full abstract
Quantum Physics
Alchemically Transferable Parameters for Variational Quantum Eigensolvers
Machine-learned circuit parameters could remove one of the main bottlenecks of variational quantum eigensolvers: the classical optimization loop, with its typical barren plateaus, local minima, and costly energy and gradient evaluations, is replaced by a direct prediction. Prior work showed that machine-learning models paired with a structured circuit design can transfer across system size and structure within hydrogenic systems. The next step: Extending them to organic molecules would seem to require re-training on use-case-specific training sets. In this work, we show that this is not necessary by establishing an alchemical approach. A geometric projection places surrogate hydrogen atoms along the bond directions of the target molecule and so maps its local bonding environment onto small hydrogenic fragments. Existing hydrogenic solvers can then be reused as they are. The only new quantity is a single projection radius per bond type, which is fitted once on a small reference molecule and reused unchanged thereafter. We demonstrate the approach on the separable-pair circuit designs that have been used in prior work and extend the approach to multi-circuit approaches that span effective bases for quantum subspaces.
Read the full abstract
Data Structures and Algorithms
Tight B-Trees for Adjacency Labeling
We show that planar graphs admit adjacency labels of size $\log n+O(\log\log n)$: every vertex gets a label of this many bits, and the adjacency of two vertices is decided from their labels alone, in constant time. This improves the $\log n+O(\sqrt{\log n\log\log n})$ of Dujmović et al. (J. ACM 2021) and the $\log n+O(\sqrt{\log n})$ of Gawrychowski and Janczewski (SOSA 2022). Equivalently, $n$-vertex planar graphs have an induced-universal graph with $n(\log n)^{O(1)}$ vertices, down from $n2^{O(\sqrt{\log n})}$. The same holds for graphs of bounded genus, apex-minor-free graphs and $k$-planar graphs. For every proper minor-closed class we get $\log n+O(\sqrt{\log n\log\log n})$, down from the $\log n+O((\log n)^{3/4})$ of Dujmović et al. (2026). Both results come from tight B-trees: nodes are split into parts of nearly equal size, so every node but the topmost stays within a factor $1\pm O(1/\log n)$ of its target size, and a root-to-leaf path loses only $O(\log\log n)$ bits.
Read the full abstract
History and Overview of Mathematics
Finite models for the order axioms of the Hilbert Plane: Redundancies and Finite Independence Models
Results on the redundancy and on the independence of the order axioms of the Hilbert plane are scattered in the literature, from Veblen's 1904 remark that the independence of the axioms of order can be shown by finite models to the redundancies removed from Hilbert's axioms of order in later editions of the Grundlagen. The aim of this note is to collect them in an organic and self-contained form for the incidence and order axioms as presented by Greenberg. In doing so we single out three redundancies inside the order axioms: the symmetry of betweenness, the existence part of trichotomy, and the second clause of plane separation. Each of them dictates which part of the corresponding axiom an independence model must violate, and we give explicit finite models of three, three, nine and twenty-five points. A model of all the incidence and order axioms, on the contrary, is necessarily infinite, so the independence of any axiom added to them requires infinite models; for two of the order axioms the finite models also give independence from all the remaining axioms of the Hilbert plane.
Read the full abstract
Numerical Analysis
A Geometry-Robust Nečas-Type Inequality for Stokes Preconditioning
Pressure mass preconditioners for the incompressible Stokes equations can deteriorate on domains with narrow gaps. We develop a pressure preconditioner that uses the distance to the boundary to account for spatially varying geometric scales. Its application consists of a pressure mass solve and a weighted pressure Poisson solve, with the squared boundary distance as the diffusion coefficient. The theoretical basis is a distance-weighted pressure norm defined by splitting the pressure into an $L^2$ component and a component controlled by its weighted gradient. We prove that this norm is bounded by the negative norm of the pressure gradient on every bounded connected open set, with a constant depending only on the dimension. A reverse bound follows from the boundary Hardy inequality, giving norm equivalence uniformly over families with bounded Hardy constants. Two-dimensional finite element experiments on slender, perforated, and composite domains demonstrate stable performance across the tested geometric parameters and mesh refinements. The smallest eigenvalues of the preconditioned pressure Schur complement remain close to $0.54$, and block preconditioning with algebraic multigrid auxiliary solves requires about 20–30 FGMRES iterations across the geometric sweeps.
Read the full abstract
Probability
Equivalence of finite and infinite survival and sharpness of the phase transition for additive particle systems
Krone's two-stage contact process is an interacting particle system in which each site on a grid can be in one of three possible states: 0 (empty), 1 (young individual), and 2 (adult individual). Only adults can multiply, and their offspring is initially young, before growing up after an exponential time. There are two death rates, one for adult trees, and a higher one for young trees. The process is additive and has an additive dual, but unlike the contact process it is not self-dual. Answering a question of Krone, for the process on the $d$-dimensional integer grid, Foxall proved that survival of finite populations is equivalent to nontriviality of the upper invariant law. Using a completely different method, we generalise this result to processes on countable groups and along the way also establish sharpness of the phase transition for a general class of additive interacting particle systems.
Read the full abstract
Astrophysics of Galaxies
MeerKAT reveals an ultra-steep spectrum radio halo in Abell 370
Hubble Frontier Field cluster Abell 370 ($z=0.375$) has been the target of a suite of legacy multi-wavelength observational campaigns and is a prime candidate to host a diffuse radio halo, given its high mass ($\sim10^{15}\,\rm{M_\odot}$) and non-relaxed dynamical state. Previous detections of its radio halo were of low significance and imaging fidelity, leading to an uncertain characterisation of its nature. We use deep MeerKAT Ultra-High Frequency observations ($580-1015\,\rm{MHz}$) to robustly map the halo's spatial extent and spectral properties. We find a radio emission size of $\sim 1.1\,\rm{Mpc}$, significantly larger than previously reported. We measure a MeerKAT UHF in-band integrated spectral index of $\alpha=-1.54\pm0.05$. This result is consistent with an ultra-steep spectrum radio halo ($\alpha\lesssim-1.5$). We find a strong, sub-linear correlation ($k=0.60\pm0.08$) between the radio and X-ray surface brightness. The radio halo's position on the $P_{1.4\,\rm{GHz}}-M_{500}$ scaling relation is underluminous, consistent with the known USSRH population. The combined evidence strongly supports that the diffuse central radio emission in Abell 370 is an Ultra-Steep Spectrum Radio Halo (USSRH), refining its previous classification and providing further support for the turbulent re-acceleration model.