Papers

The Gálvez Davison Index (GDI) is a thermodynamic diagnostic designed to summarize convective favorability in the tropical regions, yet its transferability across complex terrain and heterogeneous hydroclimates remains uncertain. Here we assess the climatological behavior and precipitation relevance of the GDI over Mexico using ERA5 hourly fields and CHIRPS daily precipitation for 1981 to 2024. We characterize the seasonal cycle and spatial contrasts of GDI and evaluate its association with rainfall using grid point and regional Pearson and Spearman correlations between daily maximum GDI and daily precipitation. To introduce an independent convective proxy, we also quantify the relationship between hourly GDI and ERA5 outgoing longwave radiation (OLR). Event level discrimination skill is assessed with receiver operating characteristic (ROC) curves and area under the curve (AUC) across regions and months. While GDI precipitation correlations are moderate and region dependent, GDI OLR coupling is consistently stronger, partly because OLR approaches a lower bound once deep convection develops, supporting its use as an intermediate proxy between thermodynamic favorability and rainfall. However, GDI values are systematically lower over elevated terrain because of the pressure level formulation of the index, which should be considered when interpreting the GDI over mountainous regions. ROC results indicate skill above random chance in nearly all region month combinations, with AUC frequently exceeding 0.75 during the core convective season. Building on these diagnostics, we propose a regionally calibrated GDI convective regime scale for May to October, adapted from the original GDI framework but recalibrated for Mexico using event conditioned distributions defined from OLR and precipitation and quantified with bootstrap uncertainty.

Span programs provide a general method for designing quantum query algorithms, and for Boolean coordinate queries their optimum cost agrees with the general adversary bound. We develop a matroidal organization of this optimization that separates dependence data determined by the query problem from the dependence structure chosen by a span program. The resulting source matroid records query dependence, target determination, and separation data, while a program matroid records dependencies among the target and response-labeled input elements of a candidate program. When a query problem admits a source matroid, we use its structure to organize adversary upper and lower bounds. For separating binary query sets in this setting, the global matroid-span-program optimum equals $\mathrm{Adv}^{\pm}_{\mathcal Q}(f)$. For a fixed regular program matroid, projective uniqueness reduces optimization over real representations to positive element weights. We express positive and negative costs as weighted basis ratios and derive exact $1$-, $2$-, and $3$-sum reductions, giving a compositional evaluation of the fixed-weight objective along a Seymour decomposition. For pointed $R_{10}$ closure, exact certificates give $3.9087553<\mathrm{Adv}^{\pm}_{\mathcal Q_R}(f_R)<3.9301$. The natural $R_{10}$ program has optimized cost $5$, while another regular program has cost at most $\sqrt{87/5}$. Recursive self-composition yields a family on $N$ input bits with randomized query complexity $\Omega(N^{0.7324867\ldots})$, an explicit quantum algorithm using $O(N^{0.6500178\ldots})$ queries, and $Q(F)=\Theta(N^\alpha)$ for $0.6204277\ldots<\alpha<0.6229063\ldots$.

Galaxy clustering and weak gravitational lensing are powerful cosmological probes, traditionally combined through galaxy clustering and galaxy-galaxy lensing in a $2\times2$pt analysis. An alternative is the foreground-background clustering cross-correlation induced by lensing magnification, but measurements have achieved lower signal-to-noise than shear-based lensing. We show that, for deep photometric samples, cancellation between galaxies with positive and negative magnification responses can suppress this signal. Using DES Year 6 photometry and Balrog synthetic-source injections, we explore high-redshift ($z\simeq0.7$-$2.5$) samples split by the sign and amplitude of their magnification response. This separation reduces cancellation and can increase both signal-to-noise and cosmological constraining power. We call the combination of foreground clustering and foreground-background cross-correlations with response-selected background samples BigMag. Forecasts with DES-like simulated data vectors show that, for an optimistic sample configuration and sufficiently precise response calibration, BigMag can achieve constraining power comparable to traditional $2\times2$pt using galaxy positions alone. The same samples also enhance inverse galaxy-galaxy lensing (IGGL), a position-shear correlation dominated by magnification when the position sample lies behind the shape sample. Combining BigMag and IGGL further self-calibrates the magnification responses and yields marginalized $S_8$ constraints comparable to traditional $3\times2$pt analyses, and tighter for the more precise calibration scenarios considered here. The framework also enables near-zero-response samples for traditional $3\times2$pt and geometric lensing-ratio analyses. These results motivate further optimization and validation of galaxy selections to realize magnification as a complementary cosmological probe. [Abrg.]

The eigenvalues of a large normal random matrix form a two-dimensional Coulomb gas. At fixed weight, the spectral droplet grows with the matrix rank, and its semiclassical density increment is harmonic measure. We freeze an $N$-eigenvalue configuration and sample $M$ further eigenvalues using the projection onto polynomials of degree less than $N+M$ that vanish at the frozen points. Their conditional law is an exact polynomial ensemble in the potential of the frozen charges. For a regular analytic droplet, we prove that the spatial distribution of the new eigenvalues converges to harmonic measure on the old boundary when $M/\log N\to\infty$ and $M/N\to0$. This convergence holds conditionally on the frozen spectrum, outside a set of frozen configurations whose probability tends to zero. The equilibrium process formed from orthogonal polynomials of degrees $N,\ldots,N+M-1$ has the same limit. An exact finite-rank identity relates polynomial moments of the conditional mean to fluctuations of the frozen spectrum, with a remainder from polynomial modes of degree $N+M$ and above.

Traditional sleep-staging methods apply the same model to every EEG epoch. Such uniform deployment expends computation on epochs that a smaller model could handle reliably, motivating cascades in which a primary classifier accepts its reliable predictions and defers the remainder to a more capable model. In this paper, we study the first stage of such a cascade: maximizing the coverage of fixed primary predictions subject to a prescribed accepted-risk target. We propose COVER (COVerage-oriented Error Ranking), which integrates two key innovations: (i) auxiliary-informed primary-error learning, which replaces maximum softmax probability (MSP) with a learned error score while preserving the primary labels, and (ii) fixed-scale scorer refinement, which builds on this score to directly maximize coverage under an empirical accepted-risk constraint rather than error-prediction accuracy over all epochs. We evaluate COVER on Sleep-EDF-20 at a 5% accepted-risk target, with subjects held out from all fitting and selection. Auxiliary-informed error learning raises mean subject coverage from 31.5% for MSP to 48.8% at similar subject-equal risk. At equal acceptance volume, with MSP accepting the same number of epochs as the learned scorer in each subject (20,639 in total), errors fall from 1,467 to 867. Fixed-scale refinement then adds 1.7 percentage points of coverage over its initialization in nested development, and COVER attains the highest mean coverage among eight evaluated scorers, 50.4% at 4.5% subject-equal risk, above the selective-ranking baseline SELE (49.3%) and the probability-fusion comparator DuoF (43.6%). To the best of our knowledge, this is the first work to combine auxiliary-informed primary-error learning with fixed-scale coverage refinement for selective sleep staging, offering a basis for reliability-aware allocation of computation in cascaded sleep staging.

Nucleon scattering off strongly deformed nuclei such as the actinides requires coupled-channels (CC) calculations in which several members of the ground-state rotational band are coupled explicitly, which makes Bayesian uncertainty quantification of deformed global optical potentials costly. We reformulate the direct boundary matching method (DBMM) by imposing its outgoing-wave boundary condition through a generalized Bloch operator; the resulting Bloch-DBMM solver converges as fast as the modern R-matrix method and extends to charged projectiles on targets with deformed charge distributions. On top of it we build a reduced-basis emulator, CCB-ROM, whose basis vectors span all channels of a $J^\pi$ block simultaneously, and generalize the coupled-channels empirical interpolation method to arbitrary non-affine deformed interactions. For $n,p+{}^{232}$Th inelastic scattering with the ground-state band coupled through the $8^+$ state, at 4 and 35 MeV, the Bloch-DBMM solver is benchmarked against Frescox, and the emulator replaces a 2700-dimensional system by one of dimension $N_b\le200$, with a cost of $\mathcal{O}(N_b^3)$ independent of the number of channels, reproducing the exact solver calculations for all five band members with speed-ups of over two orders of magnitude. CCB-ROM thus delivers differential cross sections, transmission coefficients and reaction cross sections from one calculation, at a cost compatible with Bayesian calibration of deformed optical potentials for neutrons and protons alike, and extends naturally to closed channels and to compound-nucleus reaction codes.

We study ramification invariants of finite separable extensions of valued fields that are pure on a primitive element. In the immediate unibranched case, we prove that the extended trace ideal is generated by minimal-polynomial derivatives of integral primitive elements and that the Dedekind different is its divisorial closure. For Galois extensions, we express this derivative ideal as a product of subgroup ramification ideals. These formulas extend prime-degree calculations to pure extensions of higher degree. We compute the differential module, traces, differents, ramification ideals, and character Swan ideals of a pure immediate elementary abelian extension of degree $p^2$. A totally ramified example of the same degree gives a comparison with the defectless case.

Energy Recovery Linacs (ERLs) are advanced particle accelerators that recover the kinetic energy of the beam after interaction, significantly improving efficiency. In this paper we discuss an alternative operation mode for multi-pass ERLs with an extra pass at top energy, allowing a doubled beam current at the expense of an increased energy spread. We explain the theoretical concept behind this approach, detailing how the phase shift enables an extra acceleration pass without requiring additional hardware. Preliminary simulations on a lattice similar to the PERLE one explore the feasibility of this method. The results are discussed in the context of ERL performance optimization, and potential applications for future accelerator designs are explored.

At face value, composition operators look like something first experienced in algebra. It's just composition of functions after all. However, there is a deep and technical setting in which these objects live, behaving like infinite-dimensional matrices acting on strange yet fascinating spaces. We take a novel approach to introducing a dense mathematical topic by mirroring that of a so-called scratch kitchen, where ingredients are not prepackaged, processed, or frozen but rather are fresh. While some mathematical maturity will certainly help in the same way that baking experience helps with a new recipe, nothing in one's background is assumed. Along the way, we make no reservations about making historical or philosophical comments to spark reader's interest as well. Bon appetit!

We study the minimum Vertex Cover problem under two non-adaptive offline advice models, where predictions about a fixed optimum solution are provided once as part of the input. These forms of prediction were introduced independently by Cohen-Addad, d'Orsi, Gupta, Lee, and Panigrahi (2024) and by Ghoshal, Makarychev, and Makarychev (2025). In Partial Predictions, independently revealed vertices come with correct membership labels. For every sufficiently small fixed $\varepsilon>0$, we give a randomized polynomial-time algorithm that has expected approximation ratio at most $2-(2-o(1))\frac{\log\log(1/\varepsilon)}{\log(1/\varepsilon)}$. In Noisy Predictions, every vertex instead receives a mutually independent noisy membership label of bias $\varepsilon$. For every sufficiently small fixed $\varepsilon>0$, we give a randomized polynomial-time algorithm that has expected approximation ratio at most $2-(1-o(1))\frac{\log\log(1/\varepsilon)}{\log(1/\varepsilon)}$. Our Noisy Predictions bound matches the leading asymptotic improvement below $2$ obtained by Aamand, Chen, Gollapudi, Silwal, and Wu (2025), despite using only one noisy label per vertex rather than a separate independent label for each endpoint of each incident edge. Both our bounds hold as $\varepsilon$ goes to $0$ and are strictly below $2$, the optimal approximation threshold for minimum Vertex Cover without advice under the Unique Games Conjecture, as shown by Khot and Regev (2008).

We prove the sharp horizontal isoperimetric inequality in $\mathbb H^n$, $n\ge2$, with equality precisely for left translates and Heisenberg dilations of Pansu balls, up to Haar-null sets.

For $n\ge2$, we prove a quantitative centre-rigidity estimate for bounded sets of finite Riemannian perimeter that are stationary and volume-constrained stable for normalized Riemannian perimeter with prescribed curvature $h_j$. Under confinement to a fixed compact set and cutoffs of vanishing full surface Dirichlet energy on the smooth regular boundary, the squared $L^2$ norms of the horizontal centre derivative, angle defect and weighted shape defect are bounded by $C(\varepsilon_j^2+\|h_j-2n\|_{C^2})|\Omega_j|$ when the parenthesized quantity is sufficiently small. The excess of normalized Riemannian perimeter over horizontal perimeter satisfies the same bound.

We study flows on finite volume homogeneous manifolds arising as directional restrictions of time changes of higher rank abelian unipotent actions. We establish Ratner-type rigidity results in this setting. For first row actions on finite volume quotients of $\operatorname{SL}_n(\mathbb{R})$, we show that natural quantitative assumptions on the cocycle force the flow to be algebraic and its ergodic invariant measure to be homogeneous. These results extend to higher dimensional subactions and to a general abstract setting. We also show that, on compact quotients of $\operatorname{SL}_3(\mathbb{R})$, the smooth time change is smoothly conjugate to a linear algebraic action by a leaf preserving diffeomorphism if and only if the cocycle of the directional flow has bounded deviations along one direction. We also obtain precise asymptotics for compact root increments. The principal thresholds in our results are sharp.

When documents supporting an agent's derived facts are revoked or replaced, should it repair memory or re-read current evidence? We introduce an evidence-revision evaluation on medication- and problem-list tasks from public ICU records. Under revocation, replacement and control events, we compare full and source-filtered re-reading with caching, rebuilding and graph-local repair across two 7B models. Memory is supplied in full without retrieval, and costs include ingest, revision and every use. On short records, local repair uses 5-10$\times$ fewer revision tokens than rebuilding, yet every memory pipeline costs at least twice full re-reading in held-out conditions. In a small pre-specified development sweep, adding task-ineligible documents extended records to about 10,000 tokens; at that length, memory's mean cumulative cost fell below full re-reading's after 2-14 uses, partly through truncated extraction, while source-filtered re-reading remained cheapest. In the replacement study, none of the four primary confirmatory tests reached statistical significance. These results show why the cost of agent memory after evidence revision must be assessed against source-filtered re-reading over the full pipeline.

High-accuracy static and dynamic polarizabilities of the $2^3P_u$ and $^{3}S_{g}$ states of the neutral beryllium atom are reported. The calculations employed explicitly correlated Gaussian (ECG) basis sets, and first-order wave functions were obtained by minimization of the Hylleraas functional. The final basis sets used for the response-property calculations contained c.a. 7000-8000 functions, depending on the polarizability component and the chosen symmetry. The results cover frequencies from the static limit, \(\omega = 0\), up to the third resonance for the \(2^{3}P_{u}\) state and up to the second resonance for the \(3^{3}S_{g}\) state. To the best of our knowledge, these are the first calculations of dynamic and static polarizabilities for these excited states of the beryllium atom using explicitly correlated basis sets.

For $n\ge2$, we prove that the $C^2$ boundary of a bounded nonempty open set in $\mathbb H^n$ is volume-preserving stable if and only if it is a Pansu sphere up to left translation and Heisenberg dilation. The proof combines centre reconstruction with a volume-constrained test field derived from Pansu dilation. Consequently, every isoperimetric region admitting an open representative with $C^2$ boundary is a Pansu ball up to these symmetries.

For $n\ge2$, we establish second differentiability of horizontal perimeter for compact $C^2$ boundaries in $\mathbb H^n$ and the continuous extension of their affine second variation to $C^1$ vector fields along the boundary. For the unit Pansu sphere, the Hessian agrees with the closed Jacobi form across both characteristic poles, including variations which move them. We compute the full Friedrichs spectrum of its Jacobi operator. The operator has compact resolvent, Morse index one and nullity $2n+1$; its kernel consists exactly of normalized translation speeds. Consequently, the sphere is stable under volume-preserving ambient variations.

Model-based control of an aquatic robotic platform depends on a hydrodynamic model that is costly to identify and specific to the hull, payload, and conditions it was measured in. Here, we present MOSAIC-SV, a deployable real-time adaptive dynamics identification and control system that identifies a control-sufficient dynamics model from a spec-sheet engineering prior, without dedicated identification trials, and re-estimates it at every control step of a closed-loop mission while the controller plans on it. A physically admissible unscented Kalman filter re-estimates hydrodynamic, disturbance, and actuator parameters at every control step of the closed-loop mission; while a command-dependent consider projection withholds corrections the current command cannot attribute between actuator effectiveness and external force; and a model predictive path integral controller plans on the current estimate. In simulation on a CyberShip II plant, MOSAIC-SV recovers the transit performance of the calibrated model under static mismatch and transient changes, and stays within 20% of its own transit time at the unscaled prior when its inertia or damping prior is wrong by an order of magnitude. In on-water field trials on the Blue Robotics BlueBoat, a twin-thruster catamaran, MOSAIC-SV transits at least 25% faster and predicts its own motion with at least 56% less error than its frozen engineering prior, including under an unmodeled payload. The same MOSAIC-SV system concept was also feasibly deployed on a 6.3-tonne dual outboard monohull.

Graphoid automata evaluate graphs by composing relations between words of states. We examine the equations that make this evaluation independent of the chosen graph expression, and what changes when relations are replaced by matrices over a commutative semiring. We prove that the classical graphoid equations are equivalent to the special commutative Frobenius equations together with the required symmetry laws. The known cospan and hypergraph presentations then describe equality of structural and labelled expressions, with the bone law deleting precisely the isolated internal vertices. We use established partition representations to compare this equality with equality of evaluated matrices. For any nonempty finite state set $Q$ and commutative semiring $K$, we give a uniform criterion for two structural expressions to have the same matrix. The criterion determines which isolated-vertex counts are distinguished and when a vanishing scalar makes matrices from different boundary partitions equal. For the canonical structure, the bone law holds precisely when $|Q|\cdot1_K=1_K$. Preserving all relational compositions is a separate requirement, equivalent to additive idempotency for $|Q|\geq2$. We then examine the characteristic matrices of the known relational structures on $Q$ given by disjoint unions of Abelian groups. Their Frobenius equations hold over every commutative semiring, whereas specialness depends on the order of each group and the bone law depends on the number of groups. These conditions specify which equations remain valid when weights are assigned to a graphoid automaton.

Large Language Models (LLMs) are increasingly used for automated algorithm design. However the computational cost of evaluating the generated algorithms can be excessive. We consider the common LLM-driven automated algorithm design (LLM4AD) setting in which a candidate algorithm is evaluated by aggregating its performance over a shared set of training instances. This instance-wise structure raises a natural question: must every candidate be evaluated on the entire instance set before deciding whether it remains competitive? Taking inspiration from algorithm configuration, we introduce AdaEva, a drop-in adaptive partial-evaluation framework that progressively evaluates candidates on larger subsets of the same instance pool and eliminates unpromising candidates as evidence accumulates. Importantly, AdaEva leaves the underlying LLM4AD procedure and per-instance evaluator unchanged and requires no prior knowledge about instance difficulty. We instantiate this idea using successive halving (AdaEva-S) and statistical racing (AdaEva-R), and evaluate both mechanisms across three representative LLM4AD frameworks, multiple LLM backbones, and optimization domains spanning combinatorial and continuous black-box optimization. Under matched evaluation budgets, AdaEva more reliably balances evaluation effort across candidates than fixed partial-evaluation strategies, yielding strong search efficiency and anytime performance together with improved held-out generalization across the evaluated settings.

This study develops a survey-calibrated, path-dependent agent-based microsimulation for evaluating crime and violence reduction policies in Bolivia. A synthetic population is constructed from the 2025 Household Survey and the fourth-quarter 2025 Continuous Employment Survey using official expansion weights and cross-survey donor matching. The model combines heterogeneous agents, place-based risk, criminal adaptation, institutional capacity, community organization, and multiple policy channels. Uncertainty is propagated through adaptive Gaussian-process and tree-ensemble emulators, global sensitivity analysis, tempered ensemble MCMC, and predictive meta-calibration. Simulated portfolios include focused policing, community control, Guardabosques-type incentives, social prevention, environmental interventions, justice reforms, investigation, rehabilitation, and financial-network disruption. Results are interpreted as structural policy scenarios rather than causal forecasts, with explicit attention to displacement, distributional effects, implementation capacity, and multi-objective trade-offs.

A deployed LLM agent emits tool calls, queries, and code that can be silently wrong – by the time the error surfaces, the action has run. Frontier chat APIs hide the model's token probabilities; the agent's stated confidence barely beats chance on the mistakes that matter; and resampling does not help, since frontier models are highly repetitive, reproducing the same call across samples. We recover the missing signal from a low-cost open-weight surrogate run in parallel. It reads the same context, schema, and proposed action as the agent, then scores the call from its own log-probabilities through a family of complementary readouts: teacher forcing and request-PMI weigh the likelihood of each argument value, a discriminative verdict judges the call as a whole, and tool-choice competition tests the function against its siblings. One principle says which to trust: a generative likelihood localizes wrong argument values, while the verdict catches holistically wrong calls. When the error type is unknown, an ensemble is the low-regret default. The readout is training-free, needs no access to the agent's internals, and costs one prefill pass alongside the tool call. On difficult coding tasks it reaches AUROC 0.825 where the actor's stated confidence is near chance (0.598), and the generative readouts beat it by +0.07 to +0.28 across three further actors. Against self-consistency it gains +0.14 to +0.19 on near-deterministic actors, at 1/K the cost. The signal drives two deployment modes: a real-time gate escalating the least-trustworthy calls for review (+0.05 to +0.30 accepted-action accuracy at 50% coverage), and confidence feedback, returning the tool result with the score so the agent adapts its next step – lifting task success on live-execution benchmarks (+0.119 and +0.137, p <= 1e-4) and beating a random-value control where step errors are silent (+0.078, p = 0.003).

The sign problem remains a major obstacle to extending the success of Monte Carlo methods to generic quantum many-body systems. A powerful alternative relies on cluster expansions, which provide convergent representations of free energies and related quantities, with applications ranging from tensor network contractions to quantum dynamics and circuits. Within their convergence regime, these expansions often yield deterministic polynomial-time algorithms. However, such methods require exhaustive enumeration of connected clusters, leading to a computational overhead of $(N\varepsilon^{-1})^{\mathcal{O}(\log\Delta)}$, where $N$ is the system size, $\varepsilon$ the desired precision, and $\Delta$ the interaction-graph degree. This becomes particularly severe in long-range systems, where $\Delta\equiv N$, resulting in quasi-polynomial runtimes. Here, we introduce a Markov chain algorithm that samples cluster expansions directly in the space of connected clusters, replacing deterministic enumeration by Monte Carlo sampling. Because sampling occurs in an abstract cluster space rather than a physical configuration space, the conventional exponential sign problem is absent within the convergence regime. For general polymer models, we establish an FPRAS with runtime $\tilde{\mathcal{O}}(N^4\varepsilon^{-3})$. For quantum partition functions at temperatures $\beta\leq\beta_{\mathrm{sampling}}\equiv\mathcal{O}(1)$, this improves to $\tilde{\mathcal{O}}(N^{3+1/D}\varepsilon^{-3})$, where $D$ is the lattice dimension. The framework directly applies to long-range quantum systems with interactions $\propto1/r^\alpha$, including $2$-local models with $\alpha>2D$.

The $k$-th power of a simple graph $G$, denoted $G^k$, is the graph with vertex set $V(G)$ where two vertices are adjacent if they are within distance $k$ in $G$. We investigate the zero forcing number of graph powers. Powers of graphs are much denser and generally not encompassed by existing results on zero forcing of graphs, hence their study requires a different approach. In contrast with the usual zero forcing behavior under edge deletion, we show that the zero forcing parameter (and variations of it) is monotone with respect to taking powers. We also determine the zero forcing number of powers of paths and cycles, together with upper bounds for powers of spiders and of grids. We then present spectral lower bounds on the zero forcing number of graph powers which uniquely use the spectrum of the base graph, as well as linear programming methods to compute these bounds. Finally, we study the sharpness of the derived bounds. To derive our results we use techniques ranging from graph theory, linear algebra and polynomial optimization.

Spin glasses are non-equilibrium magnetic systems with pronounced slow relaxation, ageing, and chaotic, synaptic-like properties, whose unique history-dependent dynamics have remained difficult to harness in functional devices. Two-dimensional van der Waals spin-glass magnets offer a route to overcome this limitation by allowing controlled coupling between magnetic and mechanical degrees of freedom. Here, we demonstrate the integration of a spin-glass functionality in nanoelectromechanical systems (NEMS) using thin exfoliated NbFeTe$_2$ films, which establishes a robust spin-glass state down to a few tens of nanometer-thick layers. Using cryogenic magnetic force microscopy, we reveal characteristic slow, hour-scale macroscopic magnetic relaxation, associated with glassy reorganisation of its magnetic domain clusters. Incorporation of thin NbFeTe$_2$ into membrane resonators allows the transduction of this glassy magnetic domain relaxation into strain, shifting their mechanical resonance frequency. This provides a functional nanomechanical readout of spin-glass dynamics and establishes a versatile platform for integrating frustrated magnetism into NEMS.

Analog quantum computing with neutral atom computers is rapidly evolving into a promising way to perform combinatorial optimization tasks at scale. As technology matures, exploring real use cases leveraging such unique information processors becomes warranted; one example is logistics management where vehicles are assigned routes to perform a given task. Autonomous aerial vehicles (drones) are increasingly used for surveillance over a given territory. In this paper, we propose an innovative framework integrating classical machine learning forecasts and quantum annealing to find the best combination of tours for a given surveillance criterion. The risk maps produced by machine learning are used to find possible surveillance tours, while a quantum-assisted heuristic chooses the best combination, producing a full surveillance plan. By modelling our decision problem for drone paths as a Team Orienteering Problem (TOP) with additional constraints, we construct a new heuristic using quantum annealing to find good solutions. We apply this framework to wildfire prevention, where finding tours most likely to detect a starting fire has clear benefits. We study the effect of larger quantum resources on heuristic performance and compare those results to the use of classical resources.

When the two largest eigenvalues of an entanglement or negativity spectrum meet, the gap of the corresponding entanglement or negativity Hamiltonian closes while entropies stay smooth, an event we call a temporal criticality. We ask when such criticalities occur in open fermionic systems, in which order, and whether they precede the geometric decoherence time, at which the negativity starts to fall. We compare loss and gain, which respect fermion parity only on average (weak symmetry) and keep the state Gaussian, with density dephasing, which respects it in every jump (strong symmetry) and makes the state non-Gaussian. We find that Gaussianity, rather than the symmetry, fixes the criterion for a criticality. In a Gaussian state the entanglement gap of a subsystem closes exactly when its two parity sectors carry equal weight, and the negativity gap closes exactly when this happens for the subsystem or its complement, so the negativity criticality never comes after the entanglement one. Uniform balanced loss and gain only shrink the two parities and leave both criticalities in place at every rate, for any quadratic Hamiltonian. In an open Kitaev chain, weak imbalanced loss and gain shift the criticalities by amounts set by the occupations at the cut, and at strong rates they allow a closing near the relaxation time only when an odd number of neighboring sites pass half filling. For dephasing we solve the non-Gaussian dynamics exactly at the point where the bond terms of the chain commute. Weak dephasing delays both criticalities and, away from the chain ends, moves them ahead of the parity balance, where the parity rules of Gaussian states would place them. As the dephasing rate approaches critical damping, both criticalities recede to infinite time, and at strong dephasing the negativity starts to fall long before either gap begins to shrink.

Reconstructing global sea surface pH from sparse observations is critical for monitoring ocean acidification and understanding marine carbon cycling. Traditional assimilation and inverse models are physically grounded but costly for large-scale reconstruction. Recent black-box and physics-guided AI models improve efficiency, but are mainly designed for passive tracers, where the reconstructed variable is also the transported inventory. In contrast, pH is an active carbonate tracer: it is the prediction target, while dissolved inorganic carbon (DIC) is the conserved carbon inventory. This mismatch can produce low pH error while violating carbonate closure and source-free carbon conservation. To address this, we introduce REACT, a carbon-first reconstruction framework that decouples transport, active correction, and chemical decoding. REACT transports a latent carbonate state with a conservative advection–diffusion solver, captures non-conservative carbon-cycle variations with a source module, decodes the corrected state into pH, and constrains the output through carbonate equilibrium. This design keeps pH as the target while enforcing consistency on the underlying carbon state. On simulation data, REACT reduces pH NRMSE by (14.7%) and chemical consistency error by (24.0%) over the best baseline. Cross-temporal-scale evaluations show robustness against error accumulation from coarse to fine temporal scales, and ablation studies validate the effectiveness of each component.

Upper mid-band spectrum is promising for 5G and 6G cellular networks, yet concrete and low-emissivity glass impose over 20 dB of building penetration loss above 2 GHz, degrading outdoor-to-indoor (O2I) coverage. A reconfigurable intelligent surface (RIS) on a neighboring building bypasses this blockage by focusing energy onto indoor users. This paper introduces PRIMO (Parallel RIS Min-Rate Optimization), a training-free joint optimizer of base-station precoding, RIS phase shifts, and RIS placement, validated by calibrated ray tracing in a MetroTech Commons digital twin (Brooklyn, NY) at 6.75 and 16.95 GHz. At equal physical aperture, half-wavelength spacing packs 6.25 times as many elements per array at FR3 as at FR1(C), so FR1(C) leads at low indoor-user counts while FR3 dominates at higher counts, where the additional spatial degrees of freedom suppress inter-user interference. Above a 1 bps/Hz service floor, PRIMO supports 92% more indoor users at FR3 than at FR1(C), and raises the minimum user rate by up to 90% over a fully converged alternating-optimization baseline while running about 100 times faster.

The most prominent feature in the ultra-high-energy cosmic ray (UHECR) sky is a significant excess of events from the vicinity of Centaurus A (Cen A). In this paper we investigate whether the excess may be caused by a past GRB 221009A-like event rather than Cen A itself. If an extraordinary gamma-ray burst (GRB) went off in one of the dwarf galaxies within the Cen A/M83 group, intergalactic magnetic deflections would naturally spread out and delay its cosmic ray signal, making it visible for $\sim 2\times 10^{5}$ years. Scaling this up, we model a local population of these exceptionally bright, narrow-jet transients within the \text{Greisen–Zatsepin–Kuzmin} (GZK) horizon. For a local volumetric rate of $\dot n_0\simeq4\times10^{-8}\ {\rm Mpc^{-3}\,yr^{-1}}$ and a jet half-opening angle of $0.8^\circ$, we estimate $\simeq3-7$ UHECR excesses across the sky at any given time. Varying the jet opening angle, GZK volume, and visibility time results in a broader range of excess values. It is possible to test this hypothesis by comparing the population predictions against existing and upcoming UHECR anisotropy data obtained by the Pierre Auger Observatory and the Telescope Array.