Peer Revenue Now accepting positions

Papers

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

All of arXiv

Monday 9 November

Software Engineering

Decalage: Uncovering the Gap Between Model-Generated Reasoning and Actual Code Implementation in Small Language Models

Agentic code generation often asks a model to explain its reasoning before it writes code, and a common way to make such a workflow safer is to add a security instruction to the prompt. We test what that instruction does in small language models, which are attractive because they run locally and cheaply. Four open code models of 1.5B to 3B parameters generate C++ from 166 sections of eleven public requirements documents, once with a general prompt and once with a secure prompt. The secure prompt more than doubles the share of reasoning about validation, resource handling and threats, and we find no evidence that the rise is larger for requirements with security content. The code changes only in part. Among functions with a body and reasoning, the share with a check that our lexical test recognises rises from 23% to 39%, but more than a quarter of the functions remain stubs, compile rates stay low, and half or more of the functions whose reasoning mentions a check contain no such check. Two rounds of compiler-feedback repair leave more than 80% of sections unusable, and neither cppcheck nor an exploratory ESBMC run separates the prompts on the usable code. A hand check by two annotators supports the stub measure and the direction of the reasoning result. For small models, a security instruction changes the reasoning broadly but the code only in part, so the reasoning should be checked against the code rather than trusted.
Read the full abstract

All authors Saif Uddin Mahmud, Afsana Tasnim, Ahmed Ryan, Md Rayhanur Rahman

Cryptography and Security

ShadowBench-MCP: Measuring Shadow Invocations in MCP-Style Tool Routing

LLM agents increasingly act as routers for tools, APIs, and software-supply-chain workflows, making an execution path - rather than generated text alone - a security-relevant model output. We study shadow invocation: a task-preserving request variant that causes an LLM router to select a riskier execution path, potentially skipping validation, provenance, staging, redaction, or audit controls even though the request does not ask for a malicious action. We introduce ShadowBench-MCP, a benchmark and metric suite that separates task preservation (TPSS), operational pressure (OPSS), and route-geometric drift (RGAS), and combines them in the pre-execution SODS score. In a focused policy-ablation study using a Groq-hosted qwen/qwen3-32b router, 120 routing decisions yield 96 original-versus-variant comparisons across six security-sensitive workflows. Efficiency-first routing produces a 29.2% shadow-invocation rate and 26.7% safety-tool skip rate. Policy-aware routing reduces these to 4.2% and 1.7%, respectively, corresponding to an 85.7% relative reduction in shadow invocations. Pre-execution signals distinguish shadow routes with ROC AUC up to 0.94. These results are early, single-router evidence of a policy-dependent routing-layer security effect, not a claim of universal vulnerability.
Read the full abstract

Fluid Dynamics

Thermochemical Rossby waves in rapid rotation: from swallowtail geometry to nonlinear selection

When heat and solute both destabilise a rapidly rotating fluid, do their buoyancies add? Codensity models assume so, giving a straight onset diagonal; spherical shells show curved and kinked thresholds. We study the Busse-Simitev rotating annulus, whose normal modes obey an exact complex cubic dispersion relation. Combining the Bilharz criterion with the neutral-curve envelope yields, in the rapid-rotation limit, an explicit swallowtail wave front, in Arnold's sense, that bends the codensity diagonal into smooth and then kinked thresholds as the diffusivity ratio decreases. Its cusp edges lie inside the unstable domain and its exposed kink ends at the swallowtail singularity; these are singularities of wave-number selection, without coalescence of temporal eigenvalues, and the swallowtail diffusivity ratio has analytical small- and large-Prandtl-number limits. At a kink, broad, fast thermal and narrow, slow compositional Rossby waves become neutral together; hidden arcs mark the birth and merger of unstable wave-number bands. The same geometry enters nonlinear selection: mean-scalar feedback factors through neutral-arc normals and scalar responses, so the mean-profile coexistence margin is positive at every corner of the leading front and vanishes quadratically with the kink angle as the waves merge. Complete cubic coefficients predict robust coexistence at a reference corner and a narrow bistable interval near merger. Periodic-cell simulations away from merger confirm the coexisting waves and their invasion thresholds.
Read the full abstract

Robotics

Accelerated Reinforcement Learning for Soft Robotics: Hours Instead of Years

Reinforcement learning (RL) requires large amounts of interaction data or a reliable simulator, both hard to obtain for nonlinear soft robotic hardware. Recent work shows that a time-scale separation inherent to soft robot dynamics justifies their reduction to low-dimensional invariant manifolds, known as adiabatic spectral submanifolds (aSSMs). We demonstrate that an aSSM-reduced model, a simple polynomial ordinary differential equation learned purely from pre-collected data, can serve as an RL training environment. On a pneumatically actuated soft arm, we train a trajectory-tracking policy in 3 hours on a six-dimensional aSSM-reduced model and deploy it without fine-tuning, whereas reproducing this policy training extrapolates to 3 years of hardware interaction. Our policy tracks dynamic targets with 40% lower error than otherwise identical policies trained on a 4.8-million-parameter recurrent neural network (RNN) or a 1.7-million-parameter transformer, which take 13 and 100 hours to identify, against 5 minutes for our mathematically justified 6,924-parameter model.
Read the full abstract

All authors Roshan S. Kaundinya, Hehui Zheng, Robert K. Katzschmann, George Haller

Machine Learning

Autonomous Continual Learning and Knowledge Transfer via Cortex-Inspired Neural Assemblies

Unlike the brain, artificial neural networks (ANNs) suffer from "catastrophic forgetting": complete overwriting of solutions when re-training on new functions ("concept shift") or on a disjoint domain of the current function ("domain shift"). ANNs require each batch of data to be roughly representative of the complete distribution, a typically expensive and impractical requirement. In this paper, we introduce Functional Task Networks (FTNs), a method of continual learning and knowledge transfer modeled after several properties of the mammalian neocortex. FTN is formulated as a distributed, stochastic mixture of experts in which a sparse, binary routing mask is computed at inference time using a small support sample of input-output pairs. Per-batch reconfiguration of the router during training, together with mask-overlap-filtered replay, shapes a basis of overlapping, partially trained subnetworks. Unrelated tasks route through mostly separate assemblies, related tasks overlap on shared components. On block-sequentially trained continual learning and knowledge transfer tasks FTN retains solutions nearly as well as disjoint-mask isolation and attains the lowest pre-update head-start loss among matched-architecture comparators. The learned mask-overlap geometry tracks the known concept-sharing structure where available. We show experimentally that under correlated concept shift, support-set mask reconfiguration is necessary even for retention. Under domain shift, the distributed bank with replay is sufficient to handle retention without the configurer. Forward transfer, or head-start, and backward transfer, are also contributed by support-set mask configuration.
Read the full abstract

Systems and Control

Rapid Heat Removal from Chip Cores: Near-Junction Active Cooling of Power Transistors via Electron Entropy Transport and Spatial Decoupling

Chip performance is limited by a shared bottleneck: the core produces heat faster than it can be moved out. We study the hardest case, RF power GaN HEMTs, whose hotspots exceed 5 kW/cm^2, and propose a horizontal near-junction active cooling topology built on electron entropy transport. Classical on-chip thermoelectric cooling faces a contest: Peltier absorption grows linearly with current, Joule heating with its square. In a vertical pillar, its own conduction resistance locks about half the Joule heat back toward the cold end, capping net cooling. From the Onsager framework we show the origin of this asymmetry: Peltier absorption localizes at the interface where the Seebeck coefficient jumps, whereas Joule heating fills the volume. The topology exploits this: a thermoelectric thin film lies flat on a diamond substrate beside the device drain (the electron exit), so bulk Joule heat leaves vertically into the substrate. A lateral fin model gives a closed form for the recirculation factor gamma: the vertical pillar yields gamma = 1, while the horizontal film yields gamma_max = sqrt(2) L_z/L_x, about 0.03 for L_z = 10 um, L_x = 500 um. The decoupling scales the net cooling ceiling and the critical runaway current by roughly 1/gamma. At equal resistance, the same film lifts maximum cooling from 13 mW to 500 mW, and the critical current from 0.65 A to 24 A. A time-constant and linear stability analysis with temperature-sensitive parameters defines the working window and safety margins for microsecond pulse cooling. The architecture leaves the device structure untouched; the principle, absorb at an interface and extract bulk heat vertically, applies to any chip with a hot core and offers a reproducible design path for near-junction thermal management.
Read the full abstract

Differential Geometry

Geometric Rigidity of Manifolds Admitting $m$-Modified Conformal Vector Fields

We establish certain integral inequalities for $m$-modified conformal vector fields on compact Riemannian manifolds and derive corresponding triviality and rigidity results. We further obtain integrability conditions for such vector fields and establish an extended Bourguignon-Ezin identity. We then prove that a compact Riemannian manifold admitting a non-trivial closed $m$-modified conformal vector field is isometric to a round sphere under either a constant scalar curvature and positive Ricci curvature assumption or a suitable bound on the Ricci curvature involving the first nonzero eigenvalue of the Laplace operator. As a consequence, we obtain an extension of a classical sphere characterization result of Yano and Nagano to the corresponding result for a non trivial $m$-modified conformal vector field assumed closed. Finally, we consider a complete Kähler manifold admitting a closed $m$-modified conformal vector field $V$, and show that $V$ is $m$-modified Killing and around a non-zero point of $V$ the manifold splits as the product of the complex Euclidean plane $\mathbb{C}$ and a complete Kähler manifold.
Read the full abstract

Multiagent Systems

Steering Emergent Behavior in Multi-Agent LLM Systems via Information Networks

Large language model (LLM) agents increasingly operate in groups, each pursuing an individual objective while jointly determining system performance. If an operator controls which agent observes which other agent, but not their payoffs or messages, how should observations be structured to raise collective welfare? We make the information network (who observes whom) a design object, decoupled from the fixed interaction network over which agents act. Ten identical LLM agents play repeated group-decision games; by default each observes only an anonymous summary of the group's previous-round outcome, and a manager with a fixed budget adds directed observation links that reveal peers' past actions. Across 1,080 episodes and six games spanning coordination, anti-coordination, social dilemmas, and discovery, we ask how much information to share and, at a fixed budget, who should observe whom. Disclosure creates behavioral coupling: observation acts as a contagion channel whose welfare effect depends on the alignment between individually attractive behavior and the collective objective. First, the value of information is alignment-contingent: more peer information tends to help or be neutral when individual behavior is aligned with the collective goal, but under free-riding or congestion externalities it amplifies collectively harmful behavior; in a public-goods game the group does best with no information at all. We read this sign separation as a cross-game regularity within one controlled LLM population, not a model-invariant law. Second, at a fixed budget, who observes whom matters primarily when coupling is harmful: a modular "barbell" that lets sub-populations entrench harmful behavior is reliably the worst wiring, reversing the network-reciprocity intuition from human evolutionary game theory. These results argue against a default of full transparency for agent collectives.
Read the full abstract

Information Theory

Binary Belief Propagation Fixed Points: Finiteness, Rigidity, and Topological Complexity

A fixed finite binary Ising model has only finitely many admissible complex belief propagation fixed points, even at degenerate parameter values. Families of positive fixed points obtained by varying external fields can nevertheless have nontrivial topology. We study this distinction through the external-field projection. For a fixed field, we prove finiteness by bounding the pole orders of the magnetizations on a hypothetical complex solution curve. The generic degree on the support graph then bounds the total local multiplicity. A graph has a single smooth real line as its full positive response for every choice of positive couplings, base field, and strictly positive sweep direction exactly when it contains no two vertex-disjoint connected subgraphs of cycle rank at least two. Two such subgraphs suffice to create a circular component. For a plane obtained by varying two vertex fields, connected simple bipartite cubic graphs with one uniform coupling and a universal bound on the fixed fields realize every genus. The minimum number of vertices grows logarithmically in the genus, by a matching algebraic upper bound on Betti numbers. Finally, we certify the complete field discriminant of an eighteen-vertex product model and recover its torus component's genus from the counts of distinct roots on all field cells. The three main theorems, including the logarithmic size bounds, are formally verified in Lean 4 with Mathlib.
Read the full abstract

Machine Learning

Variational Approach to Optimal IPS Estimator for Multi-logger Off-Policy Evaluation

We study off-policy evaluation (OPE) in contextual bandits with data collected from multiple logging policies. Inverse propensity scoring (IPS) is a standard approach to OPE and extends naturally to the multi-logger setting. However, as highlighted by Agarwal et al. [2017], neither balanced nor weighted IPS consistently outperforms the other in this setting. We resolve this dilemma by deriving an optimal IPS estimator with sample-dependent weights that minimize variance subject to unbiasedness. Using a variational calculus approach, we obtain closed-form optimal weights, yielding an estimator that is unbiased and achieves asymptotically optimal variance within a generalized weighted-IPS estimator class. Instead of a reward model, the estimator needs only a $K\times K$ matrix and a $K$-vector of moments, where $K$ is the number of logging policies, and we quantify how much efficiency this costs relative to the semiparametric oracle. Experiments on benchmark datasets show that our estimator generally improves on existing multi-logger IPS methods. We also extend our estimator to a doubly robust form by incorporating a conditional reward estimator. We compare the resulting DR extension with the semiparametrically efficient DR estimator of Kallus et al. [2021], both theoretically and empirically. We show that the two estimators achieve the same asymptotic variance when the conditional reward estimator is consistent, while our estimator can retain semiparametric efficiency under certain forms of reward misspecification. Empirically, the proposed DR estimator often achieves lower estimation error than the existing DR baseline.
Read the full abstract

Software Engineering

CRUSTPairs: A Cross-Language Dataset of Paired C and Rust Solutions

C-to-Rust migration is an active agenda, but no dataset pairs the two languages at the level of whole solutions to a shared problem. We present CRUSTPairs, 1857 C/Rust implementation pairs drawn from competitive programming archives, normalised, deduplicated, and stratified into difficulty tiers by zero-shot cross-language embedding similarity, with perrecord provenance and licensing. Cross-language retrieval is hard on this data, and harder than a single baseline would suggest: TF-IDF over shared identifiers reaches MRR 0.21 and so beats a pretrained code model at 0.19, which on the hardest quarter of the data scores 1.7x chance. Finetuning reaches 0.79, and lifts that hardest quarter 10x. Renaming every user identifier settles what is being learned: the finetuned models keep 96-97% of their accuracy without them, while the pretrained model keeps 66% - less than the lexical baseline does - and on structurally divergent pairs the finetuned models are better without identifiers than with them.
Read the full abstract

Artificial Intelligence

CASCoDe: A Closed-Loop Agentic Framework for Analog Sizing and Layout Co-Design

Expert analog layouts differ from auto-generated ones in density and regularity, not only in legality: experts make a succession of discrete, hierarchical physical decisions and revisit each after extracted electrical results, while most existing automated flows act on a physical search space fixed before the controller runs. We present CASCoDe, a closed-loop LLM agent framework for analog netlist-to-layout design that formalizes those expert decisions as a hierarchical discrete action space-sizing selection, structural floorplan, local layout refinement, routing-attempt selection-on which the agent runs feedback-driven tree search from a circuit netlist to a verified GDS. The agent authors the circuit-specific floorplans and scripts, and an execution layer of interface adaptations over adapted tools make every action land at physical implementation and be judged by DRC, LVS, and extracted-RC simulation. On 20 SKY130 amplifiers spanning single-, two-, and three-stage topologies, CASCoDe is the only flow that delivers a DRC- and LVS-clean routed layout when meeting design specifications for every circuit. It reduces area by 59-85% (mean 71%) against a conventional P&R flow on identical PCells, raises mean device density from 0.39 to 0.74 against APPlace, MAGICAL, ALIGN and ICPI. A layout that beats every baseline in area and legality appears after 11 minutes of agent time on average, and ablations show that the structured action space and the feedback-guided tree search each reduce area and search cost.
Read the full abstract

Robotics

Data-Driven Precision Control of Soft Helicoid Robots

A purely data-driven method grounded in the theory of adiabatic spectral submanifolds (aSSMs) was recently used to model and control nonlinear soft robotic simulators. Here, we extend the methodology to learn aSSM-reduced models from data collected from hardware soft robotic platforms. Specifically, we learn a five dimensional (5D) aSSM-reduced nonlinear ordinary differential equation (ODE) that accurately models the dynamics of a soft-helicoid trunk robot. We then use this ODE in a model predictive control scheme to demonstrate real-time precision control across various challenging closed-loop trajectory-tracking tasks, ranging from 2D to 3D trajectories with static and dynamic constraints. We find that our methodology enhances accuracy and precision across all control tasks compared to available data-driven linear methods and model-free reinforcement learning. While we focus here on a soft-helicoid trunk robot, our methodology relies solely on hardware data and is therefore applicable to a broader class of soft robotic systems.
Read the full abstract

All authors Roshan S. Kaundinya, Mark Leone, Roberto Sabella, Hugo Buurmeijer, Luis Pabon, John Irvin Alora, Marco Pavone, George Haller

Algebraic Geometry

Derived invariance of holomorphic two-forms and homological motives of stably holomorphically symplectic varieties

Let $\Phi:D^b(X)\to D^b(Y)$ be a derived equivalence between smooth projective complex varieties, and let $h=c_1(L)$ for an ample line bundle $L$ on $X$. For $z\in\mathbb C$, transporting the cup-product action and evaluating at $e^{zh}$ gives a homomorphism of graded algebras $\Theta^\bullet_{F,z}:H^\bullet(Y,\mathcal O_Y)\to H^\bullet(X,\mathcal O_X)$, where $F$ is the cohomological transform of $\Phi$. We prove that $\Theta^\bullet_{F,i}$ is an isomorphism in degrees one and two. In particular, $X$ and $Y$ have the same number of holomorphic two-forms. The proof in degree two uses the Hodge-Riemann bilinear relations on $X$ and on $X\times X$. We also prove that being stably holomorphically symplectic is a derived invariant, where a smooth projective variety is called stably holomorphically symplectic if it is a direct factor of a smooth projective holomorphic symplectic variety. For derived-equivalent varieties in this class, we construct an isomorphism of rational homological motives as unital algebra objects. In particular, their graded rational cohomology algebras are isomorphic compatibly with the Hodge structures.
Read the full abstract

Chemical Physics

Integer electron-count cuts in variational second-order reduced-density-matrix theory. II. A convergent hierarchy of rigorous relaxations for real-space pair-density representability

Variational second-order reduced-density-matrix calculations can dissociate molecules into fractionally charged fragments; Part I largely removed this failure with integer electron-count conditions on fixed partitions. Building on the realizability theory of point processes, we place these conditions in a convergent hierarchy of rigorous relaxations, defined by finite occupation linear programs (LPs) on spatial partitions, for real-space pair densities at fixed one-particle density. Every level contains all representable pair densities, the levels tighten under nested refinement, and their intersection characterizes pair densities admitting a symmetric $N$-particle probability-measure extension. In a fixed orbital basis the levels give monotonically increasing lower bounds to the full configuration-interaction energy, which they need not reach, even for the Coulomb repulsion. The spin-summed real-space cuts of Part I are its two-cell members. The LP optimum is also a measure: at fixed one-particle density it equals the $L^1$ distance of the coarse-grained candidate from representability and increases under refinement to that of the candidate, with a vanishing coarse-graining error; its primal law constructs a representable approximation, and its dual a violated valid inequality, used as a cut in molecular calculations. The count conditions are not implied by the usual positivity conditions: moment data that satisfy all three-particle positivity conditions, including $T_1$ and $T_2$, give through Gaussian orbitals a smooth pair density that violates an integer count. For atoms at fractional electron number, ensemble relaxations with the integer-count condition lie closer to the exact piecewise-linear energy than the zero-variance fixed-$N$ form.
Read the full abstract

High Energy Physics (Theory)

Landau-Ginzburg Theory of a Holographic Superconductor: A Horizon Perspective on the Zero-Temperature Ground State

We investigate the effect of a quartic scalar self-interaction, $\lambda|\phi|^4$, on a holographic superconductor in which non-linear electrodynamic couplings generate a superconducting dome containing an extended Lifshitz regime. The quartic interaction leaves the linearized scalar dynamics around the normal phase unchanged while modifying the non-linear solutions where the scalar field is large. As $\lambda$ increases, the solutions within the dome pass through three stages. The Lifshitz region shrinks until its scaling solution disappears at $\lambda^*$, leaving a no-AdS region, beyond whose edge the flows end before reaching any asymptotic region, and above $\lambda_c$ the dome consists entirely of asymptotically AdS hairy black holes. At the bottom of the dome we identify the point where the horizon value of the scalar field is largest, which we refer to as the critical point. There this value lies at the deepest symmetry-breaking minimum that the horizon effective potential, of Landau-Ginzburg form, can have. Unlike a conventional Landau-Ginzburg description, this interpretation rests on the scalar field at the horizon rather than on the boundary order parameter. Below $\lambda^*$, the constant scalar of the Lifshitz fixed point sits instead at the minimum of the effective potential of its asymptotic scaling region.
Read the full abstract

Analysis of PDEs

Double exponential growth for the Euler equation on the whole plane

We consider the incompressible Euler equation on the whole plane $\R^2$, without boundary and without forcing, and we construct two smooth solutions whose vorticity gradient grows fast for all large times. The first solution has finite energy, integrable vorticity and initial data in every Sobolev space. Given any function $\Gamma(t)$ with $\Gamma(t)/t\to 0$, the logarithm of its vorticity gradient is not bounded by any multiple of $\Gamma(t)$, and, at every sufficiently large time $t$, it is larger than any multiple of $t/\log_*t$, where $\log_*$ is the iterated logarithm. The second solution has bounded vorticity, bounded vorticity derivatives on every finite time interval, and a velocity given by the absolutely convergent Biot–Savart integral which grows at most logarithmically in space. Its vorticity gradient grows double exponentially in time for all large times. Its initial vorticity is neither integrable nor square integrable, and its initial velocity and strain are unbounded. Both constructions place infinitely many localized amplification mechanisms far apart from each other in one initial datum. The first one is based on the hyperbolic stagnation point behind a smooth traveling dipole, controlled for long times by an energy–Casimir argument. The second one is based on the hyperbolic point of the Bahouri–Chemin cross on the torus, transported to the plane by spatial dilations and by the uniqueness theory of Serfati solutions.
Read the full abstract

Machine Learning

Scalable Counterfactual Explanations for Time Series via Policy Learning

Generating counterfactual (CF) explanations for time series is challenging due to complex temporal dependencies and the computational inefficiency of traditional instance-specific optimization. We propose a policy learning framework for counterfactual explanation in time series, trained as a one-step contextual-bandit policy-learning problem. By training an actor-critic agent in the latent space of a pre-trained autoencoder, our method learns a transferable policy that captures global structural regularities. Unlike existing methods that rely on iterative optimization for each new input, the learned policy enables direct generation of valid and plausible counterfactuals through a single forward pass, avoiding per-instance optimization at inference time. Experiments on 13 UCR benchmark datasets show that our framework achieves high validity and strong plausibility across multiple anomaly detection metrics. Overall, by decoupling counterfactual generation from per-instance optimization, the proposed method provides an efficient and scalable alternative to existing approaches for time series counterfactual explanations.
Read the full abstract

All authors Peiyu Li, Pouya Hosseinzadeh, Soukaina Filali Boubrahimi, Shah Muhammad Hamdi

Sunday 8 November

Probability

Stationary causal enhancements of fractional Brownian motion at small Hurst indices

We construct stationary, complete-past causal geometric enhancements of prescribed multidimensional fractional Brownian motion for every $0<H\leq1/4$. They retain the first level, are exactly self-similar, and commute with linear coordinate maps. The construction removes a remote-past cutoff from Volterra signatures. At reciprocal indices $H=1/N$, a lagged top-degree logarithmic-velocity subtraction yields a quantitative joint cutoff and zero-lag limit with sharp moment bounds. Centered velocities and white-noise localization provide the critical cancellation. The limit is stable under weak concentration at zero of deterministic mass-one signed lag profiles on a common bounded lag range, with uniformly bounded total variation. Top-central perturbations give nonuniqueness at nonreciprocal indices, preserving all lower levels and symmetries. At reciprocal indices, square-integrable whole-line causal extensions with fixed lower levels and common-noise shift and scaling identities differ by a deterministic central drift. Logarithmic centering removes this drift. The enhancements yield controlled integration, rough differential equations, a finite-word regularity-structure model and an adapted realization by ordinary signatures.
Read the full abstract

General Relativity and Quantum Cosmology

Probing the Graviton Mass through the Stochastic Gravitational-Wave Background Last-Scattering Surface

A nonzero graviton mass makes gravitational waves of different momenta travel different distances during the same cosmic interval. A relic stochastic gravitational-wave background (SGWB) with a finite momentum bandwidth therefore samples a shell of finite thickness, rather than the thin source shell of massless propagation from a single epoch. We describe this effective last-scattering surface by mapping the momentum and emission-time distribution into a radial visibility window in an FLRW background. The calculation uses the geometric-optics propagation relation and includes both dispersion and finite emission duration, without a small-mass or narrow-band approximation. We then calculate how radial averaging changes the angular power for an early-produced background whose momentum channels trace a common source-fluctuation field. For a scale-invariant source spectrum, moving a thin shell leaves its angular power unchanged, so the reduction in power isolates the effect of its finite thickness. For a logarithmic momentum width of $0.5$, mass-to-momentum ratios $q=0.3,1,3$ give angular-power ratios of approximately $0.974$, $0.771$, and $0.508$ at $\ell=10$. At a representative frequency of $3\,\mathrm{nHz}$, these ratios correspond to masses of approximately $3.6\times10^{-24}$–$1.2\times10^{-23}\,\mathrm{eV}$. We establish a framework that maps the graviton mass to the SGWB angular power spectrum through its effect on propagation distances and the radial window, for a specified source distribution and frequency selection. Using this framework for graviton-mass inference requires a source and detector model; the calculation here establishes the angular response rather than an observational sensitivity.
Read the full abstract

All authors Hongguang Zhang, Zhengyang Qiu, Sirui Wang, Jiayi Wei, Jing Zhang

High Energy Astrophysical Phenomena

Direct inference of cosmological mass growth via timing of black hole - pulsar binaries

Cosmologically coupled black holes (CCBHs) respond to the expansion of the Universe, independently of conventional astrophysical processes. Recent CCBH solutions in General Relativity demonstrate evolution of gravitating mass $m_{\rm BH}\propto a^\kappa$, where $a$ is the cosmological scale factor. This can be tested through the orbital period timing of black hole (BH) pulsar (PSR) binaries. In contrast to previous methods used to constrain $\kappa$, pulsar timing measurement baselines span decades instead of gigayears, allowing sources of bias to be well-controlled. As a test case, we analyze the globular cluster BH+PSR candidate J0514–4002E (spin $P\sim 5.6~{\rm ms}$, orbital period $P_{\rm b}\sim 7~\rm days$) under the assumption that the companion is a CCBH. Taking the difference between the observed orbital and spin period derivatives removes all kinematic effects, leaving only the pulsar's intrinsic spin-down $\dot{P}_{\rm int}$ as unknown. Setting $\dot{P}_{\rm int} = 0$, its physical minimum, gives a robust upper bound $\kappa \leqslant 3.95$ at $95\%$ confidence. We estimate $\dot{P}_{\rm int}$ through the population of millisecond pulsars matched in spin and infer $\kappa = 0.82 \pm 1.67$ and $0.75 \pm 1.54$, using $H_0 = 67.6$ and $73.5\,{\rm km\,s^{-1}\,Mpc^{-1}}$, respectively. Adopting typical baseline scaling of timing precision, the value $\kappa = 3$, argued elsewhere to link energized vacuum BH interiors to accelerated cosmological expansion, can be excluded or detected for this $2.35^{+0.20}_{-0.18}\,M_\odot$ companion at $3\sigma$ and $5\sigma$ confidence by 2031 and 2036, respectively.
Read the full abstract

All authors Kevin S. Croker, Gibson B. Bowling, Christopher Cain, Jake Summers, Duncan Farrah, Jeremy Coles, Hayley Williams, Chris Pearson, Joshua Grumski-Flores, Viktor T. Toth, Joshua Cohon, Rogier A. Windhorst, Leander Thiele, Gregory Tarlé, Gustavo Niz, James W. Rohlf, Nicolas Fernandez

Hardware Architecture

Multicore Quantum Architecture for Scalable and Energy-Aware Green IoT Computing

The rapid growth of the Internet of Things (IoT) is increasing the computational demand placed on edge and backend processing infrastructures, motivating scalable architectures that can reduce execution and communication overhead. This paper presents a communication-aware comparison of homogeneous multicore classical and quantum architectures for Green IoT using the Qcomm simulator. In addition, a mesh-based, homogeneous, quantum-classical multicore architecture is proposed to enable scalable, efficient execution of distributed quantum workloads. We evaluate a 4x4 mesh of 16 processing cores under regular, classical-bottleneck, and quantum-bottleneck conditions, and study the effects of traffic pattern and Light-to-Matter (LTM) port provisioning on performance. Results show that butterfly traffic yields the largest gains over uniform traffic in both execution time and qubit coherence, while performance benefits from additional LTM ports saturate beyond four ports. At the algorithm level, homogeneous Multi Quantum Processing Units (QPUs) achieve 10x, 9x, and 7x energy efficiency improvements in comparison to Multi-CPUs for Quantum Approximate Optimization Algorithm (QAOA), Variational Quantum Eigensolver (VQE), and Quantum Neural Network (QNN), respectively. The paper's findings show that communication-aware multicore quantum architectures are promising for latency-sensitive IoT processing. Furthermore, the results validate the proposed architecture and provide a foundation for future scheduling and resource management techniques to improve scalability and efficiency in multi-QPU systems.
Read the full abstract

All authors Seham Al Abdul Wahid, Farah Mohammadi, Yazan Otoum, Arghavan Asad

General Physics

The Mechanics of Curvature Inflation

We examine slow roll directly in the Jordan-frame Hubble equation. For the quadratic Starobinsky action, this equation has an exact mechanical form. Comparing nearby solutions at the same Hubble value gives a local contraction law on the expanding branch with decreasing Hubble rate. A signed square root of the Hubble parameter also continues the vacuum background smoothly through its Hubble zeroes. For metric actions built from a smooth function of the Ricci scalar, coefficients depending only on the Hubble parameter require a derivative expansion. We derive the terms through cubic order and test both the equation residual and the error accumulated by an independently evolved approximation. A regularized inverse-logarithmic extension provides a nontrivial example. On the parameter slice studied, we prove a positive lower bound on the second curvature derivative of the action and compare separate Jordan- and Einstein-frame integrations. The third-order approximation reproduces the exact inflationary trajectory more accurately than the second-order approximation over the interval tested. We also compute the linear scalar and tensor spectra with numerical convergence checks. The benchmark gives a scalar spectral index of 0.97384 and a tensor-to-scalar ratio of 0.00527, evaluated 60 exact Einstein-frame e-folds before the first acceleration exit. These values depend on the chosen action and reference event. The mechanical form supplies approximation tests for the background; the scalaron and its perturbations follow from the original gravitational action.
Read the full abstract

General Mathematics

A Las Vegas algorithm for satisfiable SAT formulas, with an application to timetabling

Many scheduling problems need many different feasible timetables, and a view of how much freedom each event has. We study a Las Vegas algorithm that samples satisfying assignments of a Boolean formula one variable at a time: it draws each value from an estimate of its conditional probability, propagates, and restarts on a contradiction. Every output is a checked solution and comes with the exact probability of producing it. We prove that a run succeeds with probability at least $2^{-K}$, where $K$ is the expected Kullback–Leibler divergence between the true and estimated conditional probabilities along the run of a uniform solution. We also bound the distance of the outputs from uniform from the samples alone; it is a divergence in the opposite direction to $K$, so a sampler can succeed more often while being farther from uniform. On generated timetabling instances counted exactly, belief propagation on the timetabling problem brings $K$ to at most about one bit, and the estimated counts are within 2% of the exact ones. On a real university instance of the International Timetabling Competition 2019, rerunning only the failed runs delivered all 72 requested timetables with each of two estimators. Belief propagation succeeds in 31% of runs. Steered by a Markov chain that carries a nearly feasible timetable, it succeeds in 65% of runs, with outputs at least 300 bits from uniform. There the importance weights are too uneven to estimate the number of timetables, and we certify a lower bound instead. For random 3-SAT, an estimator whose $K$ grows logarithmically would give a polynomial-time Las Vegas algorithm beyond the densities at which any algorithm is proved to succeed; we conjecture that one exists below the satisfiability threshold.
Read the full abstract

Social and Information Networks

Triadic and assortative structure of absorbing states in the Axelrod model

We study the Axelrod model of cultural dissemination, viewed as a finite-state Markov chain on an arbitrary finite graph, and classify its absorbing states. An absorbing configuration splits the edges into interior edges, whose endpoints share their whole culture, and boundary edges, whose endpoints share nothing. A partition of the vertices into connected blocks is the domain partition of some absorbing configuration exactly when its quotient graph is properly $q$-colourable. The configurations with a given domain partition correspond to homomorphisms from the quotient into the $F$-fold categorical power of $K_q$, so their total number is a sum of chromatic polynomials over connected partitions; this gives closed forms for trees and cycles and the value $2^{n+F-1}$ for $q=2$ on connected graphs. In absorbing configurations no triangle carries exactly one boundary edge, the open-wedge mass of the interior graph is at most that of the host graph, and Newman's assortativity coefficients for domains, cultures and single features are ordered. The feature counts are martingales and their sums of squares submartingales, which yields an absorption-time bound of order $n^2$ that is attained on paths. Exhaustive enumeration and simulations illustrate the results.
Read the full abstract

Computer Vision

Editable Video Context: Mid-Context KV-Cache Reuse for Context-Managing Video LLMs

Context Language Models let a model edit its own context, while Suffix Cache Reuse makes edits to text efficient by adjusting the positions of cached keys for surviving tokens. We extend these ideas to video language models with multimodal rotary position embeddings (M-RoPE). We represent a long video as timestamped segments. After viewing a sparsely sampled version, the model selects segments for decoding at 4x the frame rate, and the remaining segments are removed from its context. Our rebuild method reuses the KV cache of surviving segments by rotating their keys with a single offset shared by the temporal, height, and width components of Qwen2.5-VL's M-RoPE. A content-hashed cache also avoids repeating vision-encoder computation for those segments. In preliminary experiments with Qwen2.5-VL-7B on synthetic needle-in-a-video-haystack questions over public-domain films, the method reduces prefill tokens and end-to-end rebuild latency relative to full recomputation while usually producing the same answer. Densifying the segment containing the answer improves accuracy when that segment is known. When the model selects segments itself, accuracy does not improve: it rarely identifies a needle that was absent from the sparse sample. We release the code, raw logs, and scripts used to reproduce the paper's results.
Read the full abstract

Computer Vision

LVRE: A Local Visual Reasoning and Editing Agent for Optical Chemical Structure Recognition

Optical chemical structure recognition (OCSR) demands strict all-or-nothing graph fidelity, where a single misidentified bond order or formal charge makes the predicted graph differ from the reference structure. While vision-language models (VLMs) can reinspect ambiguous visual regions, unconstrained full-graph regeneration often introduces regressive errors into previously correct substructures. We present LVRE, a training-free visual agent that refines molecular graphs via targeted visual inspection and reversible transactional editing. Rather than regenerating the entire graph sequence, LVRE examines a high-resolution crop alongside the global diagram and local graph neighborhood, proposing modifications to named graph records that a deterministic controller checks. A controller executes each proposed modification as an isolated transaction, and a visual review module arbitrates which edits to commit or roll back. Across four foundation VLMs, LVRE improves graph accuracy by 8.0-8.5 percentage points over single-pass recognition and by 3.0-3.3 points over fixed local review. In a paired evaluation of 1,000 diagrams using Gemini 3.8 Flash, LVRE achieves 100 successful repairs while introducing only 15 regressive errors, in contrast to global reflection which yields 65 repairs against 30 introduced errors. On this paired set, LVRE introduces half as many errors on initially correct predictions as global reflection.
Read the full abstract

General Relativity and Quantum Cosmology

Holographic Einstein rings of a covariant quantum corrected AdS black hole

We investigate holographic Einstein rings in a covariant quantum corrected Schwarzschild–AdS$_4$ spacetime using a massless scalar probe. At fixed horizon radius $r_h$, the deformation parameter $\zeta$ leaves the metric mass parameter, coordinate surface gravity, and photon sphere radius unchanged, while increasing the photon potential maximum and the asymptotic lapse coefficient $A=1+\zeta^2$. The critical coordinate impact parameter approaches $A^{-1/2}$ for large black holes, whereas the corresponding incidence angle in the normalized boundary frame approaches $\pi/2$. Ray tracing and local WKB barrier data recover the same geometric scale. Quasinormal spectra computed with AdS boundary conditions provide a comparison at finite multipole number. These results separate the finite radius effect of the deformation from the change in boundary time normalization.
Read the full abstract

Computer Vision

T2T Technical Report: Synthesizing GUI Agent Trajectories at Scale from Video Tutorials

Training generalist GUI agents requires large volumes of interaction trajectories comprising screenshot sequences paired with grounded UI actions, yet collecting such data through human demonstration is expensive and difficult to scale. Although public software tutorial videos contain rich operational demonstrations, their continuous pixel streams lack discrete action boundaries, semantic intents, and precise spatial coordinates, making them difficult to use directly for model training. To address this, we present T2T (Tutorial-to-Trajectory), a fully automated pipeline for synthesizing GUI agent trajectories at scale. T2T spans desktop, web, and mobile platforms to convert unannotated video tutorials into structured, executable interaction trajectories. To guarantee trajectory quality, T2T combines active segment filtering, keyframe identification via localized UI change, shortcut-free observation frame selection, presence-aware grounding, and two-stage action-space normalization. These refined trajectories are then converted into multi-task training views spanning navigation, planning, forward dynamics, and grounding. We fine-tune Qwen3-VL-8B on a subset of the more than one million human-annotation-free training samples generated by this pipeline. Evaluations show that the model achieves consistent gains across multiple benchmarks covering desktop, web, and mobile environments, including OSWorld-G, ScreenSpot-Pro, AndroidControl, and AgentNetBench, indicating improved core capabilities in grounding, step-level instruction execution, and trajectory-level decision-making. Notably, all gains are achieved without any in-distribution training data from target domains. Learning exclusively from trajectories extracted from public videos via the T2T pipeline, the model exhibits cross-application and cross-task generalization capabilities.
Read the full abstract

All authors Qingsong Xie, Yanlin Fei, Nazhou Liu, Jiacheng Tang, Madalina Ciobanu, Navan Preet Singh, Qingqing Mao, Ritankar Das

Mathematical Physics

Algebra-Valued Yang-Mills-Type Theories: Polynomial Obstructions and Holonomy Reductions

We study algebra-valued Yang-Mills-type ($A$-YMT) theories on a smooth $n$-manifold $M$, where connections take values in an $\mathfrak{m}$-graded algebra $(A, *)$ and descend to an associated bundle $E_A = P \times_\rho A \to M$. We derive the field equations, gauge transformations, and generalized Bianchi identities, showing that affine Ehresmann compatibility forces the vertical image of $\mathfrak{g}$ into the Lie nucleus $\mathcal{N}_{\mathrm{Lie}}(A)$ of $[x,y]_* = x*y - y*x$, and that gauge covariance and the Bianchi identity are obstructed precisely by the Jacobiator $\mathrm{Jac}_{[\cdot,\cdot]_*}$ (leaving principal $G$-gauge symmetries anomaly-free). We then prove obstructions in three settings. First, internal-product Chern-Weil and self-dual $A$-YMT functionals vanish in weak $(2,s)$-solvable algebras, invariant under $(Q,S)$-splitting extensions. Second, in reductive Cartan-Yang-Mills theories on $A = \mathbb{R}^k \oplus \mathfrak{h}$, quartic-block parenthesization yields a strict cutoff at $n \ge 8$, while binary-tree halving yields an exact Jacobsthal-number cutoff ($n - 4 \notin \bigcup_{r \ge 3}\{2 J_r, J_{r+1}\}$, plus rank vanishing for $n \ge k + 5$). We catalog Berger holonomy reductions for $4 \le n \le 7$, proving an on-shell vanishing (torsion-free) vs. off-shell survival (torsionful) dichotomy in 4D for $U(2)$ and $SU(2)$ connections—via $e \curlywedge_\rtimes (e \curlywedge_\rtimes \Omega_\omega) = \frac{1}{2}(e \curlywedge_\rtimes e) \curlywedge_\rtimes \Omega_\omega$—and in 7D for $G_2$-connections. Third, we prove an associator-quotient collapse and Jacobiator vertex dichotomy for odd-dimensional internal $A$-Chern-Simons theories, with applications to $A$-$BF$ theories, octonionic and reductive Higgs Proca mass spectra, and reductive dualities.
Read the full abstract