Peer Revenue Publish or profit

Papers

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

All of arXiv

Saturday 5 September

General Relativity and Quantum Cosmology

Joint Constraints on Quantum Oppenheimer–Snyder Black Holes from High-Frequency Quasi-Periodic Oscillations in X-ray Binary Systems and the S2 Star Orbit

We study a quantum-corrected version of the Oppenheimer–Snyder (OS) spacetime in which classical collapse is modified by a single effective-parameter \(\alpha\), representing small-quantum-gravitational effects in a phenomenologically rescaled-form. Although the correction enters through a term proportional to \(\alpha M^2/r^4\), it produces noticeable changes in the strong-field region while leaving weak-field physics almost unchanged. One of the main results is the appearance of a minimum-mass for horizon-formation, \(M_{\min} \sim(0.5-1.2)\,M_\odot\) for the effective-parameter \(\alpha\approx0.22\). We emphasize that this \(M_{\min}\) is expressed in terms of the effective-parameter \(\alpha\); in terms of the bare loop-quantum-gravity (LQG) parameter, \(M_{\min}\sim M_{\text{Pl}}\sim10^{-5}\)~g, consistent with the theoretical-expectation. Orbital properties are also slightly modified, with the innermost-stable-circular-orbit (ISCO) shifting to \(r_{\rm ISCO}=6M-\alpha/(12M)\), giving corrections-of-order \( \sim 10^{-4} \) for stellar-black-holes and \(\sim 10^{-10}\) for Sgr A$^\ast$. In the same region, radial epicyclic frequencies change by about \(5\%-10\%\) near \(r\sim5M\), while vertical modes are less affected, leading to small but structured shifts in quasi-periodic oscillation (QPO) behavior. Using four X-ray binaries (GRS 1915+105, XTE J1550-564, XTE J1859+226, GRO J1655-40), we find a consistent range \(\alpha=0.22\pm0.10\), with masses between \(5.4-12.4\,M_\odot\) and emission radii \(r/M \sim 5.5 - 8.6\). The observed QPO frequencies, lying in the range \(100-450\) Hz, are well reproduced within this framework. Weak-field tests such as S2 and Mercury still allow \(f_{\rm sp}=1.10 \pm 0.19\), leaving room for these small strong-gravity corrections.
Read the full abstract

All authors A. Errehymy, Y. Khedif, M. Daoud, B. Turimov, S. Usanov, F. Turaev, Z. Yasakov

Multiagent Systems

An Evolutionary Computation Framework for Multi-Agent Q-Learning with Mean-Field Environmental Feedback

Multi-agent reinforcement learning in networked populations is governed by the interaction between individual adaptation, local encounters, and changing environmental conditions. To study this interaction, we formulate a coupled learning–environment model in which agents update stateless $Q$-values on a fixed graph, while their population-average behavior drives an environmental variable that dynamically modifies the payoff matrix. Under a first-order mean-field closure, we derive a deterministic transport equation for the population distribution of $Q$-values and couple it with a projected discrete update for the environmental state. The resulting model is evaluated against finite-network Monte Carlo simulations on random regular, Erdős–Rényi, Barabási–Albert, and random geometric graphs. Across the tested parameter ranges, the mean-field system reproduces the main macroscopic cooperation and environmental trajectories, and the trajectory-level root-mean-square error generally decreases with population size and average degree. The analysis further shows that environmental feedback reshapes the learned action-value ordering, while reinforcing feedback can produce pronounced dependence on the initial learning bias and resource level. The environmental timescale also plays an important role: a rapid response can drive the resource state to a boundary before learning adapts, whereas a slower response preserves the interaction between behavioral learning and environmental recovery. These results provide a population-level description of coupled reinforcement learning and environmental dynamics and characterize the performance of the mean-field approximation within the tested network and parameter ranges.
Read the full abstract

Atomic Physics

Opacity predictions in plasmas under stellar conditions using deep learning

The aim of this work is to predict the opacity of plasmas under stellar conditions. We focus on iron and nickel, as these elements have been extensively investigated both theoretically and experimentally. In certain regimes, notably under non-local thermodynamic equilibrium, calculating the spectral opacity can be computationally demanding. To mitigate this difficulty, we employ deep learning models to predict accurate opacities while achieving a substantial reduction in computational cost. Specifically, we develop and train a hybrid model that combines a convolutional neural network with a multilayer perceptron. The proposed approach provides both the spectral opacity over the 0–10,000 eV energy range, and the corresponding mean opacities for large sets of temperatures and mass densities, in close agreement with standard computational methods. The agreement is very satisfactory, except in the L- and M-shell energy ranges for nickel at high temperatures, where a very large number of emerging transitions leads to substantial discrepancies.
Read the full abstract

Computer Vision

Preserving Subject-Clarity in Image Outpainting with Multiscale Wavelet Supervision

Commercial and advertising images are frequently affected by poor framing, partially cropped subjects, truncated text or logos, and insufficient context, all of which can reduce subject clarity, i.e., the ability of an image to clearly communicate its primary subject. Image outpainting offers a scalable solution by extending image boundaries and recovering missing content and context. However, existing diffusion-based outpainting methods often produce visually plausible completions while degrading subject fidelity through structural inconsistencies, semantic drift, or loss of fine-grained detail. To address this limitation, we propose a subject clarity outpainting framework that combines vision-language model (VLM)-guided semantic conditioning with multiscale wavelet supervision for subject-localized detail preservation. To support training, we develop a subject-centric data curation pipeline that constructs subject-intersecting outpainting pairs from advertising and natural images. The resulting objective introduces no additional inference cost and is designed to be compatible with diffusion-based backbones. Across four advertising and natural-image benchmarks, our method improves subject clarity, reducing subject-centered DreamSim error and FID on average by 3.0% and 2.4% over matched supervised fine-tuning, and by 10.8% and 7.7% over the strongest state-of-the-art approach per dataset, respectively.
Read the full abstract

All authors Abhilash Neog, Taewan Kim, Yi Wu, Xu Chen, Jian Jiao

Populations and Evolution

On the formulation and analysis of a stage-structured eco-evolutionary model with pulsed disturbances

Motivated by the pulsed control and resulting evolution of harmful biological populations, we develop and analyze an ordinary differential equation model of an n-dimensional stage-structured population undergoing periodic disturbances. The eco-evolutionary model couples both ecological dynamics through life cycle information, and evolutionary dynamics of allele frequency changes. Building off prior work in eco-evolutionary models and theory of linear systems, we show that if a resistant allele is present in the population, the population will always approach full resistance. Furthermore, we establish persistence conditions and show that these depend solely on system dynamics at full resistance. Finally, we demonstrate the utility of this model via an application to the Spotted-Winged Drosophila, an invasive species often treated periodically with insecticides. We use a parameterized model to further understand the complexities of relationships between ecological and evolutionary features of a species and its resistance evolution due to pulsed disturbances.
Read the full abstract

Social and Information Networks

Collective Hysteresis and Multistability in Threshold Networks

In a mechanistic model of the dawn chorus, Kaye showed that heterogeneous activation thresholds and a shared feedback signal determined by the population's active fraction can produce abrupt collective activation and hysteresis. We extend this mechanism to a network of interacting agents. Each node has a continuous activation level, and a nonnegative row-stochastic matrix determines how node activities contribute to individual feedback. We prove that sufficiently weak feedback yields a unique globally attracting equilibrium. For all feedback strengths, the homogeneous dynamics exactly reproduce Kaye's scalar equation; consequently, network topology does not alter the folds or cusp of the homogeneous branch, and no heterogeneous mode becomes unstable before the homogeneous mode. For equitable partitions, the network admits an exact quotient system in which nodes within a block receive the same aggregate input from every block. When blocks are uncoupled, the quotient reduces to independent copies of Kaye's scalar equation. We show that every assignment of stable scalar equilibria to blocks persists under sufficiently weak interblock coupling, producing a combinatorial family of stable quotient equilibria that lift to stable full-network equilibria and remain under small perturbations that break exact equitability. For two symmetrically coupled blocks, branches with unequal block activities terminate at a pair of symmetry-related cusp bifurcations. Near the onset of bistability, we derive scaling laws for the interblock coupling at which these bifurcations occur, the activity difference between the blocks at bifurcation, and the corresponding shift of the external stimulus from the scalar cusp. Numerical continuation confirms the scaling laws for gamma, logistic, and normal threshold distributions.
Read the full abstract

History and Overview of Mathematics

A machine-checked proof of the Dong-Yang classification of optimal (n,4) binary codes for BSCs

We present a machine-checked Lean~4 formalization of Dong and Yang's classification of optimal finite-length $(n,4)$ binary block codes for binary symmetric channels. The formalization was developed mainly by feeding the paper's proofs to an AI tool. To establish correctness, the authors verified the main theorem statements in Lean and the accepted axioms. This note discusses the corrections and simplifications made to the AI-generated formalization, and records discrepancies found in the paper during the formalization. The Lean code is available at https://github.com/shhyang/n4code_lean.
Read the full abstract

General Mathematics

Consequences of Matrices Sharing Eigenvalues and Eigenvectors

While studying for a linear algebra final, the first named author prepared some test questions for herself to see how well she understood the material, and asked the second named author: \emph{If $A$ and $A^T$ have the same eigenvalues and eigenvectors, is $A$ a symmetric matrix?} We show how this excellent question is a great springboard to related questions, in particular when do equal eigenvalues and eigenvectors imply the matrices are, if not equal, at least closely related (such as similar or the transpose/complex conjugate transpose of each other)? The answer depends on how we interpret the question, and provides a great opportunity to talk about creating good questions. In particular, we characterize matrices $A$ for which the transpose or conjugate transpose shares the same eigenvectors (regardless of eigenvalues) and, for each eigenvalue, the same eigenpair (equivalently, the same eigenspace). Thus, a square matrix $A$ is Hermitian if and only if $A^*$ has the same eigenpairs as $A$; moreover, if $A$ is a real matrix with real eigenvalues and $A^T$ has the same eigenvectors as $A$, then $A$ is symmetric.
Read the full abstract

General Mathematics

New Generalizations of Two Ramanujan Series for $1/\pi$

By utilizing two known hypergeometric summation identities, we extend two \mbox{well-known} Ramanujan series for $1/\pi$ by making each of them a member of an infinite family of series. We also find the corresponding families involving harmonic numbers and odd harmonic numbers. To further illustrate our method we derive families of \mbox{Ramanujan-like} series for $1/\pi$ associated with a hypergeometric series derived by Lavoie, two hypergeometric series from Bailey's book and a recent hypergeometric series found by Campbell. Thus, by leveraging known hypergeometric summation identities, our approach bypasses the more tedious, traditional limits or complex Fourier-Legendre expansions; offering a cleaner, unified template for generating Ramanujan-like series for $1/\pi$.
Read the full abstract

Hardware Architecture

Shift-Accumulate Attention: Multiplier-Free Query–Key Products for Transformer Decoding

Power-of-two (PoT) quantisation turns a multiplication into a bit shift, so far only for the post-softmax attention–value product. The earlier and larger product, S=QK^T, has not been reformulated the same way. We quantise the key cache to a signed power-of-two fixed-point code, so that every scalar multiplication in QK^T becomes a sign flip, a bit shift and an integer accumulation. A fixed-point head-room F>= e_max turns every shift index r=F-e into a non-negative left shift, making the accumulation exact in integer arithmetic. We add a sub-power mantissa extension that halves the score error for one extra shift-add, a shift-exact online softmax whose running maximum lives in an integer log2 domain so that every rescaling is itself an exact shift, and a log-quantised AV product. Fused CUDA kernels in a 1.1B Llama decoder on an RTX 4090: at B=8, T=32k the 4-bit kernel is 4.60x faster than FP16 scaled dot-product attention with a 2.5x smaller KV cache, and batched decoding crosses over at B=64 to 1.22x FP16 throughput at 11.02 against 13.77 GB peak memory. A matched INT8 multiply–accumulate kernel on the same tiling reaches 5.29x: on a GPU with a hardware 4-way INT8 dot product, shift-accumulate is not faster than the MAC it replaces. An iso-storage control–an 8-bit shift code of exactly one byte per key, run through the identical kernel–prices the arithmetic substitution alone at 2.02x at comparable accuracy. The advantage of the PoT representation is therefore its density and the removal of the multiplier, not raw GPU throughput. A sweep over exponent width, mantissa levels, granularity and rounding gives the design lesson: nearest-PoT relative error is scale-free, so extra exponent bits buy nothing (eps_S=0.1302 at 3, 4 and 5 bits) and accuracy must come from mantissa terms.
Read the full abstract

Geometric Topology

The real quotient of the Cartwright-Steger surface

Borisov and Yeung showed that the Cartwright-Steger surface $X$ is defined over ${\mathbb {Q}}$, hence complex conjugation gives an antiholomorphic involution on $X$. We show that the orbit space $Y$ is a closed oriented smooth four-manifold homeomorphic to ${\mathbb {CP}}^2\# (S^1 \times S^3)$, presenting $X$ as a double branched cover $X\to Y$, with branch locus diffeomorphic to the connected sum of three copies of the real projective plane.
Read the full abstract

Computer Vision

MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads

Multimodal Large Language Models (MLLMs) often struggle with hallucinations, thus hindering their reliable practical applications. Existing attention-based mitigation methods mainly rely on indirect signals (e.g., attention weights) that fail to accurately reflect the actual information shift underlying hallucination generation. In this paper, we propose HEAL, Head-lEvel information disentAnglement and caLibration for identifying and mitigating hallucinations. HEAL first employs causal noise intervention on multi-head outputs to filter out causally redundant heads. Subsequently, it disentangles information distribution within the remaining heads via the counterfactual Difference-in-Differences, categorizing heads into four types. Through analysis, we observe: hallucinations happen when information distribution drifts away from a healthy equilibrium in synergy heads, not strongly correlated with the quantity or strength of modality-specific heads. Motivated by this insight, HEAL injects dynamic information calibration factors into the value vectors of synergy heads and actively regulates visual-language dependencies, steering the output distribution towards factual evidence. Extensive experiments demonstrate that HEAL effectively reduces hallucinations across multiple MLLMs, offering a simple and interpretable pathway to enhance model trustworthiness.
Read the full abstract

All authors Meng'en Qin, Junye Chen, Jucheng Liu, Yinchen Liu, Youlu Xing, Song Wang, Ruize Han

General Mathematics

Fourfold Is Enough: Connected LEGO Models from One Brick Type

Can a LEGO model built from many kinds of rectangular bricks be rebuilt using only the familiar $2\times4$ brick? Enlarging every dimension by four makes the pieces fit, but they must also connect. We give a four-layer arrangement that joins every finite face-connected union of lattice cubes, using eight upright bricks per source cube, and show that no smaller universal enlargement works. A second question has an unexpected answer: if connectivity is set aside, the minimum integer scale is always $1$, $2$, or $4$, never $3$. The missing value follows from a checkerboard argument applied separately inside each enlarged cell.
Read the full abstract

General Mathematics

Counting Type I and Type II solutions of the Erdos-Straus equation: concentration, local obstructions, and a pointwise comparison

For a prime p >= 5 let f_I(p) and f_II(p) count the Type I and Type II solutions of the Erdos-Straus equation 4/p = 1/x + 1/y + 1/z, in the sense of Elsholtz and Tao. We do not address the Erdos-Straus conjecture itself; our subject is the number of solutions. We turn a bijection of Bradford into an exact counting identity: f_I(p) and f_II(p) are sums, over the denominators n coprime to p, of the number of divisors of n^2 in a single residue class modulo e_n = 4n - p. A dual identity puts both counts on the same integer and the same modulus, distinguished only by the classes -p and -1. Our main unconditional result holds for every numerator m: the number of Type II solutions of m/p = 1/x + 1/(pY) + 1/(pZ) with Y <= K is at most sum_{n <= K} tau(n) = K log K + O(K), uniformly in p and m; on average over p <= N it is only O_m((log K)^3 log log K). Hence Type II solutions concentrate at x/p -> 1/m, while Type I solutions cannot. The numerator carries arithmetic information in two independent ways. A real character chi modulo e_n that is trivial on the primes dividing n annihilates Type II when chi(-1) = -1, and Type I when chi(-1)chi(m_0) = -1, where m_0 is the squarefree kernel of m; these coincide exactly when m is a square, so for the numerator 4 no criterion modulo squares can separate the two types. Separately, f_I(n) = f_II(n) = 0 for every odd perfect square n whenever 4 divides m, generalising a theorem of Elsholtz and Tao. Finally we prove the sharp local inequality 2B(n) >= A(n) for e_n in {1,3,5}, with a construction showing it fails for every odd e with 7 <= e <= 2*10^4; we conjecture f_I(p) > f_II(p) for all p > 5 with p not congruent to 1 mod 8, verify this for p <= 10^6, and reduce it to a finite window. We also locate exactly the factor log log N in the Elsholtz-Tao upper bound for the sum of f_I(p).
Read the full abstract

Artificial Intelligence

OpenDiscoveryTrace: Process Traces for Evaluating AI Scientist Workflows

Existing benchmarks for autonomous AI scientists evaluate only final outputs—generated code, hypotheses, or papers—yet discard the reasoning process by which those outputs were obtained. This makes it impossible to audit scientific methodology, diagnose failure modes, or distinguish systematic reasoning from fortunate guessing. We present OpenDiscoveryTrace, a public dataset of 558 complete AI scientific agent trajectories that captures how models reason, not just what they produce. Each trajectory records a structured 9-field-per-step trace—including thoughts, tool calls, observations, errors, revision triggers, and self-reported confidence—as models execute 124 scientific tasks spanning drug discovery, materials science, genomics, and scientific literature analysis. The dataset covers seven models: three frontier models (GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro; 124 trajectories each, fully balanced across domains and difficulty levels) and four open-weight models (Qwen2.5-7B, Mistral-7B-v0.3, Phi-3.5-mini, and Qwen2.5-1.5B; 30 each), plus 60 live-retrieval variant trajectories. Pilot analysis on 363 LLM-judged trajectories reveals that process traces expose behavioral differences invisible to output-only evaluation: all three frontier models achieve comparable success rates (84–89%), yet Claude Opus 4.6 produces 30$\times$ more errors than GPT-5.4 (2.5 vs. 0.08 per trajectory, $p < 0.0001$, Cliff's $\delta = 0.613$), with qualitatively different error profiles—66.7% tool misuse for Claude versus 83.6% reasoning errors for GPT-5.4. We define five benchmark tasks with baselines from logistic regression, random forests, LSTMs, and Transformer models. The dataset, trace schema, agent harness, and benchmark definitions are publicly available under CC BY 4.0 to support research on process-level evaluation, scientific agent auditing, and AI governance.
Read the full abstract

High Energy Physics (Phenomenology)

On the analytic continuation and matching of threshold singularities from Vacuum Amplitudes

We provide a detailed discussion for all kinematical configurations of the analytic continuation to negative values of initial-state on-shell energies used to generate interferences of scattering amplitudes from vacuum amplitudes in the Loop-Tree Duality. We also extend previous discussions on the matching of threshold singularities and introduce angular averaging as an efficient strategy to achieve a local cancellation when the matching of threshold singularities from vacuum amplitudes is constrained by flavour.
Read the full abstract

All authors Selomit Ramírez-Uribe, Prasanna K. Dhani, German F.R. Sborlini, Germán Rodrigo

Rings and Algebras

Homomorphisms and subalgebras of free convex semilattices

A convex semilattices is a convex algebra endowed with an additional semilattice operation that satisfies a distributive law. We investigate free convex semilattices by studying the homomorphisms into products R{\Omega} and, more specifically, the homomorphisms between two free finitely generated convex semilattices. We characterise the finitely generated subalgebras of a free finitely generated convex semilattice.
Read the full abstract

Probability

A Branching Process with Geometric Emigration and the "Have Your Cookie and Eat It" Random Walk

We consider a random walk on the integers that at each site is biased to step right until the first time it has stepped left from that site, thereafter stepping left or right from that site with equal chance, introduced by Pinsky. Pinsky calculated the speed of the walk and the probability that such a walk escapes to infinity for some ranges of \(p\) and conjectured these formulas hold over a wider range. To study this walk, we introduce a branching process featuring geometric emigration and answer some questions concerning its life-periods, transience and recurrence, and the mean of its stationary distribution when positive recurrent. We apply these results to confirm Pinsky's predictions. For another application, we calculate the speed and escape probability of a related self-interacting random walk.
Read the full abstract

Dynamical Systems

On the topological dynamical system of tilings

Tilings can be studied from the perspective of dynamical systems and ergodic theory. Some fundamental features of tiling spaces as topological dynamical systems, such as the continuity of the action by translation, the compactness with the tiling metric, and the minimality of the system generated by a repetitive tiling, are among the most frequently mentioned basic facts in the dynamical theory of aperiodic order. Nonetheless, despite their classical status, complete proofs of these facts, together with the preliminary material they rely on, to the best of our knowledge, are not available in the literature. The purpose of this survey is to give a self-contained, rigorous, and pedagogical treatment of these foundations with detailed preliminary material. We hope that this exposition will provide a useful reference and contribute to a clearer and more accessible understanding of the fundamentals of the theory of tiling dynamical systems.
Read the full abstract

Number Theory

Density one for lattice point visibility along polynomials with at least two distinct roots

We prove that every nonzero integer polynomial with at least two distinct complex roots has lattice point visibility density one, resolving the generalized form of the Visibility Density Conjecture for nonzero polynomials. This extends the origin-passing case established by Chaubey, Pandey, and Regavim. Visibility is taken along the curves $y=tF(x)$ with rational $t$, with a point visible if no positive lattice point on the same curve has a smaller horizontal coordinate. The proof uses a greatest common divisor cutoff to reduce the problem to finitely many equations of the form $F(b)=qF(a)$, with $0<q<1$; for each equation, the positive integers $a$ admitting a positive integer solution $b<a$ form a set of density zero. This elementary argument removes the proper-power hypothesis of earlier work without requiring estimates uniform in the ratio.
Read the full abstract

General Economics

Housing Price Appreciation, Housing Affordability, and the Spatial Restructuring of Toronto Commuting

This study investigates the spatial relationships among housing-market appreciation, housing affordability, and automobile commuting to Toronto across 23 census metropolitan areas and municipalities in southern Ontario from 2016 to 2021. Utilizing Statistics Canada Census data, the analysis employed linear and quadratic ordinary least-squares regression, Global and Local Moran's I, and residual spatial diagnostics to assess geographic patterns in average dwelling values, owner shelter-cost burden, and lone-driver commuting. The study found that average dwelling values increased significantly throughout the region, maintaining a strong nonlinear relationship with distance from Toronto, which indicates a persistent metropolitan accessibility gradient. Commuter volumes also demonstrated pronounced nonlinear distance gradients; however, housing appreciation was not significantly correlated with percentage changes in commuting ($R^2$ = .045, p = .331), suggesting that the expansion of the housing market did not result in proportional growth in Toronto-bound automobile travel. The shelter-cost burden exhibited weaker distance relationships and greater residual spatial dependence, indicating localized affordability influences not fully explained by distance or dwelling values. Overall, the findings suggest an increasingly extensive regional housing market with spatially heterogeneous affordability and commuting outcomes, highlighting the importance of integrating housing, transportation, and regional planning.
Read the full abstract

Quantum Physics

Hierarchical Fourier Approximation for Variational Quantum Distribution Learning

We study variational quantum distribution learning through a hierarchy of Walsh–Fourier approximations on the Boolean cube. At each level, a selected set of target Fourier coefficients defines a spectral truncation, which is projected onto the probability simplex and used as the target of a quantum circuit Born machine. Parameters learned at one level initialize the next through a warm-start map. We prove an end-to-end expected learning guarantee where the approximation term is determined by the omitted Fourier mass, while a normalized unbiased estimator yields an explicit statistical bound for empirical truncations. We then instantiate the abstract discrepancy conditions for total variation distance and relate the resulting distributional error to quantum-state fidelity. The total-variation specialization incurs the explicit factor $2^{n-1}$ under our normalized $\ell_2$ convention and is therefore informative only for sufficiently concentrated Fourier tails. The framework does not establish global trainability or eliminate barren plateaus; rather, it identifies the conditions under which low-to-high spectral training admits a approximation–estimation–optimization analysis.
Read the full abstract

Cryptography and Security

MARS: Detecting Unauthorized Variable Manipulations in Multi-Application PLC Runtimes

Programmable Logic Controllers (PLCs) increasingly run multiple applications alongside the main control program, with shared access to PLC variables. Yet, Industrial Control System (ICS) defenses primarily detect malicious updates by checking whether variable values violate expected bounds, without considering which application performed the update. A malicious application can exploit this gap by modifying variables within normal bounds while still driving the physical process toward an unsafe state. Even when such manipulation is detected, operators cannot identify the responsible application because PLCs do not associate variable updates with application identity. We present MARS, an automated framework for application-level authorization and attribution of PLC variable manipulations. MARS profiles applications on an isolated virtual PLC (vPLC) to derive application-specific variable-access policies and uses a shadow vPLC during operation to attribute production-PLC updates to individual applications without instrumenting the production controller. MARS also detects manipulations that occur only on the production PLC and therefore have no corresponding update on the shadow vPLC. We evaluate MARS on manufacturing, chemical, and water-treatment systems against attacks in which unauthorized applications manipulate PLC variables while remaining within normal bounds. Our results show that MARS detects these manipulations and identifies the responsible application.
Read the full abstract

K-Theory and Homology

The K-theory of uniform Roe algebras for coarse structures generated by finite-rank free abelian subgroups

For a uniformly locally finite coarse space $X$, the uniform Roe algebra $C_u^*(X)$ is the operator norm closure of the controlled operators on $\ell^2(X)$. The $K$-theory of uniform Roe algebras is known in asymptotic dimension zero, but it is not fully understood in higher dimensions. We compute $K_0(C_u^*(G,\mathcal E))$ and $K_1(C_u^*(G,\mathcal E))$ for every countable discrete abelian group $G$ and every finite-rank free abelian subgroup $H\leq G$, where $\mathcal E$ is the coarse structure generated by $H$. We use the Proietti–Yamashita spectral sequence to express the $K$-theory in terms of $H_*(H;\ell^\infty(G,\mathbb Z))$, which we then compute.
Read the full abstract

Rings and Algebras

Cohomology theory of Novikov algebras and applications

In this paper, first we give a new characterization of the cohomology of pre-Lie algebras using the Chevalley-Eilenberg cohomology associated to a morphism from the operad of Lie algebras to Hadamard product of the operad of pre-Lie algebras and its Koszul dual operad. Then we apply the same approach to study the cohomology of Novikov algebras, and give the cochain complex explicitly. The cochain complex of the underlying pre-Lie algebra is shown to be isomorphic to the quotient of the cochain complex of a Novikov algebra. Consequently, there is a long exact sequence connecting the cohomologies of a Novikov algebra and the underlying pre-Lie algebra. The cohomology of a Novikov algebra with coefficients in a representation is introduced using pseudo-tensor categories. As applications, we show that infinitesimal deformations and abelian extensions are classified by the second cohomology groups with different coefficients. Various examples are given to illustrate the difference between the cohomology of a Novikov algebra and that of the underlying pre-Lie algebra.
Read the full abstract

All authors Pavel Kolesnikov, Yue Li, Yunhe Sheng, Nanyan Xu

Combinatorics

A negative answer to the Erdős-Sárkőzy question

For a finite set $A$ of positive integers, let $H(A)$ be its set of subset sums, and let $g_3(n)$ be the least $N$ for which some $n$-element set $A\subseteq\{1,\ldots,N\}$ has $H(A)$ free of nonconstant three-term arithmetic progressions. Erdős and Sárkőzy asked whether $g_3(n)\gg 3^n$. We prove \[ \liminf_{n\to\infty}\frac{g_3(n)}{3^n}=0. \] More precisely, for every $\epsilon>0$ there is an integer $d\ge2$ such that $g_3(d\ell)\le \epsilon 3^{d\ell}$ for all sufficiently large $\ell$. The proof uses Korsky's characterization of the problem in terms of ternary coefficient sums and a consequence of an OpenAI construction that provides positive integer coefficients whose linear form is injective on large integer boxes. A base-three expansion then gives the result.
Read the full abstract

Computer Vision

CST-WM: A Causally Structured World Model for Embodied Visual Tracking

Embodied visual tracking requires a robot to choose actions that keep a moving target observable at a suitable distance, and to recover it after occlusion, out-of-view drift, or distractor crossings. We cast the task as planning over future target evidence with an action-conditioned world model. In logged tracking data, however, the behavior policy's actions are correlated with where the target is, so a generic predictor can learn a shortcut: it writes the current action directly into its prediction of target evidence, instead of letting the action affect that evidence only by moving the robot and changing what it observes. We call this failure causal hallucination; the resulting rollouts look plausible but rank candidate actions for the wrong reason. We propose CST-WM, a causally structured world model whose state is split into target-evidence, robot, and observation branches. Its transition removes the same-step edge from action to target evidence but keeps the path through robot motion and the resulting views, so candidate actions are still distinguished by their predicted ego-motion. With rollout-based model-predictive control, a single model handles both steady following and re-acquisition after target loss. On EVT-Bench and Habitat 3.0, covering standard tracking, target-loss recovery, and cross-dataset transfer, CST-WM improves following, distance-range control, safety, and re-acquisition over reactive trackers and world-model baselines, and removing the action mask causes the largest drop in re-acquisition among our ablations. Offline, CST-WM has lower multi-step rollout error, and its ranking of candidate actions agrees better with the simulator's. On a Unitree Go2 quadruped, CST-WM succeeds in 20 of 30 real-world trials under occlusion, distractor crossing, and fast motion, against 14 for TrackVLA.
Read the full abstract

All authors Junyi Hu, Shuaihang Yuan, Jiazhao Liang, Yi Fang

Databases

Compiling Linear Datalog to SQL for Program Analysis

Datalog is a declarative query language that has proven highly effective for expressing static program analyses. Although Datalog has deep roots in database theory, most recent advances have largely emerged from the programming languages and compiler communities, with systems such as Soufflé. In contrast, modern relational engines have made significant progress in optimizing recursive SQL. This paper revisits the connection between Datalog and relational databases, advocating recursive SQL as a backend for Datalog evaluation. We present a compilation framework that translates Datalog programs, particularly those in the Linear Datalog fragment, into equivalent recursive SQL queries. To bridge the gap between Datalog and SQL, the compiler routes every program through an intermediate language called Midlog. The compiler additionally recovers functional dependencies from the program and exposes them as schema keys, unlocking the engine's standard query optimizations. This approach enables existing database engines to execute a broad class of program analyses, outperforming the Soufflé engine by up to an order of magnitude on the Umbra backend. Umbra achieves a geometric-mean speedup of 5.46$\times$ at 8 threads, whereas DuckDB is competitive with Soufflé single-threaded and is slower at 8 threads (geometric-mean speedup of 0.68$\times$). Furthermore, the generated SQL is portable; it runs on seven database systems without any engine modification. Our results highlight what the relational engines require to fully support Datalog for large-scale program analysis.
Read the full abstract

Soft Condensed Matter

Exploring causal relationships in elastomeric fracture

The conventional interpretation of elastomeric fracture treats the tearing energy G as an output determined by tear speed, taken to equal crack velocity vc, concluding that higher vc produced greater viscoelastic dissipation to increase G. Based on spatially and temporally resolved polarized optical microscopic (str-POM) measurements of tip stress tip stress, we use prenotched pure shear experiments to clarify the causality: Local crack-tip stress determines the network lifetime, making vc the passive kinetic output. In the elastic limit that can be readily achieved for a well crosslinked elastomer in a wide range of temperatures, Rivlin-Thomas scaling holds, and vc depends only on the applied strain, independent of stretch rate, indicating one-to-one correspondence between vc and tip stress. In a highly stretchable elastomer made with reduced crosslink density, vc no longer correlates well with the far-field load as rate-dependent viscoelastic processes emerge to affect network lifetime. Because of the rheological effects on how chain tension builds in the elastomeric network, vc is higher at a common nominal strain when it is imposed with higher stretch rate. confirm that different stretch rates produce different levels of tip stress. The same tip stress still produces the same vc. With str-POM measurements we are also able to elucidate the nature of temperature dependence: at a given tip stress,, crack growth is slower at lower temperatures; conversely it requires higher tip stress to produce the same vc at lower temperatures.
Read the full abstract

Classical Analysis and ODEs

Lifting curved Kakeya sets to linear Kakeya sets

We prove that many curved Kakeya problems, originally motivated by Hörmander's oscillatory integral problem, lift to the classical Kakeya problem in higher dimensions. As a consequence, if the classical Kakeya set conjecture were true in all dimensions, the families of curves whose curved Kakeya sets have full dimension are dense among Hörmander-type families. This is in contrast to the nowhere dense Bourgain's condition, which is a necessary condition for best-case curved Kakeya maximal function estimates. Furthermore the linear Kakeya set conjecture would imply a fairly complete understanding of Kakeya sets of quadratic curves, as first studied systematically by Wisewell. These results give a better understanding of a question posed by Guo–Guth–Nadjimzadah–Shen–Zhang, and they more generally show that the classical Kakeya conjecture in high dimensions rests on a broad range of curved Kakeya problems in lower dimensions.
Read the full abstract