Machine Learning
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 6 November
Uncertainty bands on wind forecasts help inform reserve sizing and commitments, but often fail to deliver their stated coverage. We apply conformal prediction to native bands to see what it adds and what it costs, comparing six forecasters, from ARIMA to a tabular foundation model, at two wind farms in Great Britain (GB) under operational conditions. We show that calibration repairs empirical marginal coverage with little or no Winkler Score penalty, but this apparent success conceals under-coverage in high winds. Calibrating by forecast wind regime turns marginal coverage into regime-level coverage, cutting day-ahead out-of-band volume in high winds by over 40% relative to pooled calibration while narrowing bands overall.
Quantum Physics
Arrival-time problem in the quantum information era - observation of a Zeno crossover in monitored arrival-time distributions
When does a quantum particle arrive at a detector? The temporal distributions of itinerant quantum particles have been debated since the birth of quantum mechanics. Here we study the arrival-time problem in periodically monitored spin chains and quantum circuits. The arrival dynamics exhibits two regimes, the quantum Zeno and the low-monitoring regime, separated by an optimal monitoring period set by the single-particle bandwidth. In support of the theory, on superconducting quantum hardware, we measure arrival-time distributions for single and multiple interacting particles in one- and two-dimensional circuits. The measured distributions resolve both regimes and agree closely with quantum-trajectory theory. The crossover persists in interacting many-particle systems: the optimal monitoring point is preserved while many-body interference reshapes the distributions. Zeno suppression likewise persists in two-dimensional geometry. Together, these results show that the two regimes are a generic feature of monitored propagation rather than a single-particle effect. To our knowledge, our work constitutes the first experimental measurement of arrival-time distributions in a monitored quantum transport setting, addressing one of the longest-standing problems in quantum mechanics.
Read the full abstract
Numerical Analysis
A virtual element exterior calculus framework
We develop a conforming virtual element framework for differential forms on general polytopal meshes in arbitrary dimension and at arbitrary order. Recursive local Hodge-Laplace boundary value problems define a trimmed virtual de Rham complex, from which moment constraints yield a complete subcomplex. The local spaces of these two complexes contain the trimmed and complete polynomial spaces of differential forms, respectively, and admit unisolvent degrees of freedom and computable polynomial projections. The canonical interpolants commute with the exterior derivative, and both complexes have cohomology isomorphic to that of the continuous de Rham complex. For both virtual families, we establish negative-norm inverse estimates and mesh-uniform norm equivalences in all form degrees. We further derive canonical interpolation error bounds and discrete Poincar'e inequalities. As an application, we obtain an inf-sup stable mixed Hodge-Laplace method with harmonic variables and optimal-order convergence under sufficient regularity. Moreover, the two conforming complexes are cochain isomorphic to their fully discrete counterparts (J. Eur. Math. Soc., 2025, doi:10.4171/JEMS/1602), and the norm equivalences provide a key analytical bridge between the conforming and fully discrete formulations.
Read the full abstract
High Energy Physics (Theory)
The Double Copy of the Doubly Spinning Black Ring
We establish an exact curved-background Weyl double copy for the doubly spinning Pomeransky–Sen'kov black ring. Its full Weyl tensor is a quadratic form in two source-free Maxwell fields: one generated by the stationary Killing vector, the other a flux-carrying magnetic field obtained by linearizing the dipole-ring gauge potential about its vacuum limit. The mixed Maxwell bilinear is essential: no spacetime-independent real or complex change of field basis removes it. We identify the inverse $S_{IJ}$ of the Weyl coefficient matrix as a matrix zeroth copy. The same matrix arises independently from a weighted limit of the inner products of two Killing vectors in a six-dimensional Ricci-flat parent and satisfies $\Box_g S_{IJ}=-\tfrac12 F_{I\,ab}F_J{}^{ab}$. The singly spinning limit recovers the diagonal two-field Weyl double copy of the Emparan–Reall ring.
Read the full abstract
Probability
Spectral Rigidity of Stochastic Operators with Heavy Tailed Perturbations
In this paper, we provide a new set of tools for establishing the number rigidity of point processes which are particularly well suited for the spectral processes of stochastic operators. These tools allow one to prove the number rigidity of a wide class of stochastic operators that had been previously inaccessible to the standard methods in the area, such as the variance of linear statistics, which require being able to precisely analyze correlations between individual points. We propose an alternative mechanism which uses a pathwise comparison between the growth rate of the unperturbed eigenvalues and the cumulative random noise which allows one to convert the probabilistic problem into a more tractable perturbation theory problem. This new approach allows us to prove the rigidity of previously inaccessible operators including those with heavy tailed perturbations or complex correlation structures. As a demonstration, we use these tools to prove the number rigidity of new classes of stochastic Schrödinger and Sturm-Liouville Operators which are resistant to analysis using the standard rigidity toolkit.
Read the full abstract
Soft Condensed Matter
Reversible Interchain Associations Dominate Plateau Elasticity in Conjugated Polymers
The modulus plateau is widely regarded as a hallmark of topological entanglement in homopolymer melts and has been used to identify entanglement in semiflexible conjugated polymers. Using regiorandom poly(3-alkylthiophenes) as a model system, we show that their plateau elasticity relaxes through an Arrhenius process decoupled from segmental and terminal dynamics, reorganizes persistently under mechanical perturbation, and varies systematically with side-chain length. These behaviors are hard to reconcile with a purely topological-entanglement picture and instead support a transient network dominated by reversible interchain associations, consistent with {\pi}-{\pi} interactions between conjugated backbone segments. The results thus demonstrate that plateau elasticity alone does not uniquely identify topological entanglement in conjugated polymers. More broadly, they identify the strength and lifetime of interchain association as molecular design parameters linking backbone packing to stress relaxation, viscoelastic properties, and processability in conjugated polymers, complementing their established role in electronic structure and charge transport.
Read the full abstract
Quantum Physics
Beyond the $n/2$ Threshold: Quantum Query Complexity for Computing Hamming Weight and Symmetric Functions
The oracle-interrogation algorithm [van Dam, FOCS '98] computes every $n$-bit Boolean function with bounded error using $n/2+O(\sqrt n)$ quantum queries. In contrast, parity can be computed exactly using $\lceil n/2\rceil$ queries, but whether an additive overhead beyond $n/2$ is necessary for a single Boolean output has remained open. In this work, we construct an explicit family of symmetric Boolean functions requiring $n/2+\Omega(n^{1/3})$ quantum queries at a fixed positive constant error. We also give an algorithm that recovers the exact Hamming weight using $n/2+O(n^{1/3})$ quantum queries with bounded error, and therefore computes every symmetric Boolean function within the same query bound. Together, these results determine the optimal additive order for Hamming weight recovery and for the worst case over symmetric Boolean functions at sufficiently small constant error.
Read the full abstract
Software Engineering
Agents on the Patch Path: Mining AI-Associated Security Remediation in GitHub Security Advisories
Autonomous coding agents are increasingly involved in software development, but their participation in real-world security remediation remains poorly understood. We investigate observable evidence of coding-agent involvement in commits referenced by reviewed GitHub Security Advisories (GHSAs), characterize the associated vulnerabilities, and examine review timing in relation to advisory disclosure routes. We develop a local-Git mining pipeline that identifies agent evidence through known agent accounts, explicit commit-message attribution, repository instruction files, and agent configuration artifacts. Linking these findings to GHSA-referenced commits yields 469 valid candidate advisories across 189 repositories, including 448 with direct attribution and 21 supported only by repository-level evidence. Claude is the most frequently observed tool family, followed by Copilot and Cursor. We further compare review timing against other advisories and historical baselines using an exact route–timing decomposition. Relative to advisories published in the same months, candidate advisories exhibit a modestly longer mean review lag, primarily associated with within-route timing rather than disclosure-route composition. Compared with a historical baseline, they exhibit substantially shorter mean review lags, with both route composition and within-route timing contributing to the difference. These findings demonstrate that coding-agent involvement is already observable across hundreds of security advisories and repositories, including high- and critical-severity vulnerabilities. By distinguishing direct attribution from repository-level adoption evidence and separating route-composition effects from within-route timing differences, we provide an empirical characterization of agent-associated remediation in the GHSA review pipeline.
Read the full abstract
Computation and Language
Compaction, decay and reinforcement in a two-tier conversational memory: a paired evaluation against retrieval-augmented generation
A conversational agent that remembers has to decide what to keep. Retrieval-augmented generation keeps the transcript and searches it; the alternative is to compact each turn into atomic facts and search those. We compare the two on LoCoMo, with both arms answering all 1,985 questions in one window, against the same stores, with the same generator and judge, and both predictions on one row, which licenses McNemar's exact test on the 512 questions where they disagree. On the four categories comparable work reports, compaction scores 71.4 per cent against 64.6, a lead of 6.8 points (p = 5.5e-08) at 34 per cent fewer prompt tokens. On all five, including the adversarial set comparable work drops, the lead falls to 2.0 and stops being significant, because the baseline leads adversarial by 14.6. That loss has a structure. Our lead falls as the reference answer demands more source wording, 9.6 points ahead at one or two reference words and 3.1 behind past eight, with token overlap moving the same way, so the judge is not the cause. Compaction keeps what a fact asserts and discards how it was said, so it wins where the answer is a value and loses where it is a phrasing. We also report four negative results with their power calculations.
Read the full abstract
Information Theory
An Efficient Synchronization Scheme for Distributed Bandits
We study the problem of synchronizing two players in a distributed multi-armed bandit setting. The probability of synchronization error is analyzed by recasting synchronization as a communication problem. Using a well-designed random coding ensemble, we derive error exponents for the maximum-likelihood decoder. We then introduce a universal decoder that achieves the same error exponents as the optimal decoder. Finally, applying these results to distributed multi-armed bandits, we show an upper bound on the regret of our synchronization scheme.
Read the full abstract
Machine Learning
Scaling Graph Similarity Learning with Graph Neural Networks
Graph similarity computation is an important task in graph mining with many real-world applications, such as chemical compound analysis, program dependence comparison, and social network analysis. The problem is computationally challenging: Graph Edit Distance (GED) and Maximum Common Subgraph (MCS), which quantify the distance between graphs and identify their largest shared subgraph, respectively, are NP-hard to compute exactly. To address these computational challenges, Graph Neural Networks (GNNs) have been applied to this problem. However, existing GNN-based approaches have primarily focused on small-scale graph datasets, while scalability to larger scales remains underexplored. In this work, we propose FiLSim, a scalable GNN-based framework, and investigate whether a simplified GNN pipeline can maintain competitive performance on standard benchmarks while scaling to large graphs. We additionally propose a sequential graph-editing strategy to construct large-scale graph datasets with reliable GED labels to evaluate scalability. The proposed framework is evaluated on four benchmark datasets and four newly created large-scale datasets, demonstrating its effectiveness and scalability. FiLSim remains accurate on graphs with more than 150{,}000 nodes, where the baseline methods run out of memory or fail to learn, and scores all test pairs within seconds on a single GPU.
Read the full abstract
High Energy Physics (Theory)
5d $\mathbb{Z}_2$ S-folds and Seiberg–Witten curves
We compute the Seiberg–Witten curves for $\mathbb{Z}_2$ S-folds of five-dimensional $\mathcal{N}=1$ rank-1 superconformal field theories. Two complementary constructions are presented. First, we implement the $\mathbb{Z}_2$ action by identifying pairs of points on the original curve, thereby mapping them to a single point on the S-folded curve. Second, motivated by arXiv:2112.14550, we tune mass parameters to special values, corresponding to the insertion of $D_4$ 7-branes. As a nontrivial consistency check, we show that the two Seiberg–Witten curves agree with each other when expressed in the Weierstrass normal form.
Read the full abstract
Artificial Intelligence
Maya a1: Do the Options Read the Model, or Does Their Position? Order Sensitivity in a Non-Autoregressive Decision Model
Laya answers typed multiple-choice questions in one forward pass by scoring a hidden state at one [MASK] marker per option. We re-engineer the checkpoint (421M parameters; a 2-layer option head with no positional tensor; a documented but unused parallel layout) and test whether the answer depends on where an option is listed. Presenting 951 Qwen-verified questions under every cyclic rotation of their options (or six random permutations for 12 options), the released checkpoint changes its answer for 25.8% [22.9-28.7] of questions (19.7% on the 605 English-language items alone); accuracy with the gold option first versus last differs by 10.1 points, and the flips concentrate where the model's own confidence is low (54% versus 8% across confidence terciles). Fine-tuning with Laya's trainer on 2,818 teacher-verified rows with option shuffling cuts the flip rate to 7.2%; the identical rows without shuffling give 11.4%, so shuffling is responsible for 4.2 points of the improvement. A confidence-weighted vote over orderings gives 69.7% against 69.1% for a single base pass at 4.1x the compute. Labels are LLM-made and the test sets are teacher-filtered. We also fine-tune with the authors' parallel layout on the same rows: flip rate 0.5%.
Read the full abstract
Software Engineering
When Instruction Files Go Stale: Documentation Debt and Hidden Cost in Coding Agents
LLM-based coding agents such as Claude Code are guided by repository instruction files (also called context files or agent configuration files) like AGENTS.md and CLAUDE.md, which are read in every agent session. As any documentation artifact, these files can fall behind the code and become stale (e.g., a test folder moves, but the file still points to the old one). In technical debt terms, such a stale instruction is documentation debt, and its interest is the cost it adds to every later agent session. Evaluations that check only task success can miss this cost, and the stale file can persist. In this study, we ask what leads agents to act on a stale instruction, how it affects their behavior and at what cost, and how they recover. We study three real-world stale instructions from active open-source repositories in depth, each a different kind of guidance: a test command, a code location, and a procedure for adding a component. Agents fix 22 maintenance tasks under three versions of each instruction (left stale, corrected to the current path, and removed), with the code and the task held fixed, and we analyze 732 trajectories of two models at the level of individual tool calls. We find that agents follow stale guidance when it is relevant to their tasks, sometimes even when correct information is available elsewhere. On the stale test command, Claude Sonnet 5.5 needed about two extra tool calls per run, a third more than with the corrected command; in the other cases, we could not detect a difference. Although agents generally recovered from the immediate failures, none repaired the underlying instruction file. Correcting the instructions eliminated operations on obsolete paths, while removing them prevented most. Our findings show that task success alone can hide the costs of documentation debt, highlighting the need to evaluate agent behavior beyond final outcomes.
Read the full abstract
High Energy Astrophysical Phenomena
Superorbital period evolution of Hercules X-1
Context. Hercules X-1 (Her X-1) is a well-studied X-ray binary exhibiting superorbital modulation of its X-ray flux. The origin of this periodicity remains debated between a counter-precessing twisted-tilted accretion disc and free precession of the neutron star. Earlier observations (1970-1990) reported stochastic fluctuations in the superorbital period, with no long-term trend. Aims. We investigated the long-term trend of the superorbital modulation using more recent, uninterrupted data from 1995-2025 to identify features that may favour or disfavour the proposed mechanisms. Methods. We probed the superorbital period using long-term light curves from five X-ray monitors, namely CGRO/BATSE, RXTE/ASM, Swift/BAT, Fermi/GBM, and MAXI/GSC, together with literature data spanning $\sim$50 yr. We employed multiple timing analysis tools to trace its evolution. Results. The uninterrupted data from the last 30 years show that the superorbital period is decreasing at a rate of $\sim -5 \times 10^{-5}$ s s$^{-1}$. The $O-C$ analysis reveals shorter-timescale variability superimposed on the overall decay; thus, a simple quadratic function cannot explain the observed $O-C$ variations. We also find a positive correlation between the mean count rate and superorbital period across different linear segments of the $O-C$ curve. Conclusions. Period variations over a few to a few tens of superorbital cycles suggest a restoring mechanism in the system. We discuss possible mechanisms responsible for the observed long-term decay and short-term variability. (Abbreviated.)
Read the full abstract
Mesoscale and Nanoscale Physics
Exciton design via topological spin textures
Excitons carry energy and information through a semiconductor without carrying charge. This makes them attractive for optoelectronics, but also hard to guide. Magnetic skyrmions, on the other hand, are moved by small electrical currents. Here we develop a theory of excitons interacting with a skyrmion in which the electron and the hole are treated as distinct constituents, each with its own spin–orbit coupling. We find that the topology of the skyrmion is transferred to the exciton only when the exchange coupling exceeds the relevant spin–orbit splitting. The exciton then experiences a full flux quantum, with opposite signs in the two valleys. The quantum metric of the spin texture creates a ring-shaped barrier that traps excitons, whereas the topological charge fixes the angular momentum of the trapped states and hence their optical selection rules. Because the trap moves with the skyrmion, excitons can be transported by current pulses, and small skyrmions accommodate a single exciton, which opens the door for potential hybrid applications in future information technology intertwining topological magnetic bits with optical excitations.
Read the full abstract
Statistics Theory
On the Order of Privacy and Smoothing in Functional Mean Estimation
Differential privacy (DP) is a leading framework for privacy protection, yet privacy mechanisms tailored to functional data analysis (FDA) remain relatively underexplored. For functional mean estimation, we investigate a fundamental question regarding the ordering of privatization and estimation by comparing two canonical pipelines: Fit-Then-Protect (FTP), which smooths trajectories before privatization, and Protect-Then-Fit (PTF), which privatizes observations at discrete time points before smoothing. When data are observed without error, we prove that FTP achieves strictly smaller mean squared error than PTF asymptotically (as the sample size increases) under the same $\mu$-Gaussian differential privacy ($\mu$-GDP) budget in both finite-dimensional Hilbert spaces and infinite-dimensional reproducing kernel Hilbert spaces (RKHS). Simulation studies support our theory and explore settings with observation error and fitting constraints, demonstrating that FTP outperforms PTF under small observation noise, whereas PTF can become preferable in high-noise settings. In addition to this comparison, we establish a novel GDP guarantee for Gaussian perturbations taking values in a Banach space, extending existing results beyond the RKHS setting.
Read the full abstract
Artificial Intelligence
Test-Time Compute Scales Better with Parallel Agent Teams
Test-time scaling improves the performance of a fixed model by spending more computation at inference. There are two common ways to spend it on agents when a verifier can score every attempt. Best-of-N sampling runs many independent agents and keeps the best result but saturates quickly. One large team of collaborating agents can score higher but stops improving as it grows. Can we get the best of both? Our method, parallel agent teams, does so by sampling many small teams in parallel and keeping the best team's result. We test it on 12 long-horizon tasks from GPU kernels and formal proofs to exploit development. With a 27B model and a budget of 100 agents, parallel teams of ten beat parallel agents on 10 of 12 tasks and beat one team of 100 on all four tasks where we ran such large teams. At equal dollar cost the same parallel teams beat a frontier model on half the tasks. A simple statistical model shows that individual agents lose because their many samples have a low mean and that one large team loses because a single sample gets no gain from variance. Sampling many small teams in parallel balances a high mean with the gain from variance. Parallel agent teams thus offer a more effective way to scale test-time compute.
Read the full abstract
High Energy Physics (Experiment)
Kinematic Reconstruction of Top-Quark Pairs Using Analytical and Machine-Learning Methods with CMS Open Data
The reconstruction of the top-quark-pair ($t\bar{t}$) kinematics in the dilepton $t\bar{t}$ decay channel is challenging due to the presence of two undetected neutrinos. In this work, existing kinematic reconstruction methods used by the CMS Collaboration are studied using a Monte Carlo sample with $t\bar{t}$ events from CMS Open Data at a centre-of-mass energy of 7 TeV. An improved version of the loose kinematic reconstruction, LKRv2, is developed by imposing more stringent requirements on the jet-lepton pairing entering the reconstruction. A further modification, LKRv3, exploits the kinematic similarity between the dilepton and neutrino-pair systems and improves the resolution of the reconstructed $t\bar{t}$ invariant mass relative to the baseline LKR by $7\%$, with smaller improvements for the transverse momentum and rapidity of $t\bar{t}$. In addition, two machine-learning (ML) based approaches are developed using a residual multilayer perceptron and a conditional normalizing flow. Relative to LKRv3, they improve the resolution of the invariant mass, transverse momentum and rapidity by approximately $30\%$, $20\%$ and $15\%$, respectively. In particular, a resolution of 30 GeV is achieved for the reconstructed invariant mass of $t\bar{t}$ near the production threshold. A dedicated generator-level study and a generator-to-detector transfer test demonstrate that the ML approaches learn a genuine physics-driven correction to the $\nu\bar{\nu}$ approximation, rather than merely exploiting detector-specific effects. These results demonstrate the strong potential of ML-based approaches for improving the reconstruction of dilepton $t\bar{t}$ events, while identifying the neutrino-pair approximation as an important limitation on the achievable resolution.
Read the full abstract
Systems and Control
Generalized H matrix stability for power systems
The small-signal stability of a power system model with arbitrary loads is investigated, with a focus on locally constant power loads in lossless networks. A central element of the proposed analysis is the matrix H, which represents the partial derivatives of generator powers with respect to their relative angles. The local stability of any equilibrium point is determined by the properties of H if the loads consume locally constant power. Specifically, an equilibrium point is asymptotically stable if H is positive definite and unstable if H is full-rank and not positive definite, while several quantitative relationships between the eigenvalues of H and the system Jacobian matrix are also established. The main contribution is that loads are allowed to be arbitrarily located within the transmission network, thereby generalizing the existing H matrix stability criterion which is valid only if loads are located at the same buses as the generators. The advantage of the proposed criterion is demonstrated through case studies involving systems with up to 2869 buses. Accurate stability assessments are maintained, where alternative methods, such as the NRPS model, are found to be overly conservative. Additionally, a computational acceleration by a factor of 340 relative to Jacobian-based methods is achieved.
Read the full abstract
Computer Vision
AR-Edit: Training-Free Autoregressive Video Editing without Inversion
Editing a video stream in real time, without access to future frames, is essential for interactive applications. Existing training-free video editing methods, however, assume full offline access to the entire sequence, while streaming approaches either rely on additional training or are limited to specific editing techniques. We show that this limitation can be removed by analyzing the self-attention dynamics of DMD-distilled autoregressive (AR) video diffusion models. We uncover a key asymmetry: queries and keys remain stable and aligned with their source counterparts across denoising timesteps, while values diverge and encode the evolving content. Inherited from the non-distilled bidirectional teacher, this asymmetry is strongest at high noise, where few-step streaming models operate. The resulting mechanism is simple: a single forward pass suffices to capture source identity in the value features, which can be reused throughout generation. Starting from pure Gaussian noise and injecting only these source values reconstructs the input with fidelity matching or surpassing inversion-based methods, effectively eliminating inversion. Based on this insight, we introduce AR-Edit, a training-free, inversion-free method that caches source values once and selectively injects them during denoising. Across multiple AR backbones, AR-Edit enables real-time streaming editing with superior source preservation and tonal fidelity at minimal overhead.
Read the full abstract
Robotics
Seeing What Execution Needs: Benchmarking Task-Conditioned Actionable Spatial Belief for Manipulation
When should a robot look at the cup, and when at the portafilter? A view that is informative for one task can be useless for another, even at the same workstation. We introduce Seeing What Execution Needs, a benchmark for task-conditioned actionable spatial belief in simulated tabletop manipulation. Its aligned three-domain track contains 1,080 frozen cases (360 each for coffee workstation, drawer/cabinet, and microwave/appliance). The primary coffee protocols – closed-loop actionability and same-object functional out-of-distribution generalization – produce over 115,000 evaluator episodes across evaluated methods, budgets, and inference repeats. Each case pairs candidate RGB views with a manipulation command, shared scripted skills, and hidden postconditions, directly measuring whether selected observations support execution. A metadata-assisted diagnostic reference pipeline maintains spatial-functional beliefs and ranks views by missing evidence. At primary budget settings, it reaches 98.8% closed-loop and 98.9% functional-OOD success; removing functional fields or command binding reduces functional-OOD success to 0.0% and 19.7%, respectively. Under a separate full-preview contract, seven vision-language models reach 26-43%. Replacing coverage metadata with frozen RGB-derived scores retains 98.5%/98.9% success; shuffling these scores reduces success to 15%/11%, supporting dependence on image content. Separately, the aligned three-domain evaluation at H=3 supports portability of the benchmark interface under frozen domain-specific schemas. These results suggest that spatial belief for manipulation should be evaluated by whether acquired evidence meets the current command's execution requirements.
Read the full abstract
Analysis of PDEs
On the Principal Eigenvalue of the $(p,A)$-Laplacian with Sign-Changing Morrey Potentials: Complete Continuity, Optimization, and Differentiability
We establish a unified framework to study the continuity and complete continuity of the principal eigenvalue functional $\lambda_1(V,m)$ associated with the following $(p,A)$-Laplacian operator involving both sign-changing potential $V$, and positive weight $m$ in a Morrey space: \begin{align*} -\Delta_{p,A}(u) + V|u|^{p-2}u &= \lambda m|u|^{p-2}u \quad \text{in}\,\, \Omega, u &= 0 \quad\quad\quad \quad\quad \text{on}\,\, \partial\Omega, \end{align*} where $1<p<N$, $\lambda\in\mathbb{R}$ is a parameter, and $\Omega \subset \mathbb{R}^N (N \geq 2)$ is a bounded domain with boundary $\partial \Omega$. As a consequence, we investigate the existence and uniqueness of optimal principal eigenvalues $\lambda_1(V,m)$ over a weakly compact subset of the Morrey space. Finally, we derive a formula for the derivative of $\lambda_1(V,m)$ with respect to regular perturbations of potential $V$, and positive weight $m$ within a closed subspace of Morrey space. Our results develop analytic tools to overcome the lack of separability and reflexivity in Morrey spaces to investigate this eigenvalue problem. In particular, the study is new even for the linear case $p=2$. We also emphasize that the tools established are applicable for a wide class of PDEs with coefficients in Morrey spaces.
Read the full abstract
Artificial Intelligence
Knowing When to Intervene: Causal Evaluation of AI Agents
Current evaluations of agentic AI systems primarily rely on benchmarks. However, the static nature of benchmarks makes them uninformative about the value of auditing interventions at deployment time. For reliable deployment, we often need to determine within an agent's trajectory whether or not to intervene, e.g., by verifying the agent's current output or steering it toward different behavior, using either human review or more costly models or time-consuming agentic supervision. The value of such interventions is fundamentally a causal inference question, asking whether the causal effect of such an intervention is beneficial for the final outcome. Estimating this causal effect is particularly challenging due to the dynamic nature of the problem: the impact of an intervention depends on the current context and environment, as well as the agent's future behavior. We develop the first methodologically grounded framework for evaluating the causal effects of randomized interventions within agent trajectories. By leveraging a relatively small dataset of agent trajectories collected under randomized interventions such as verification or screening, we can estimate the dynamic causal effect of a rich class of (contextual) intervention policies. The structural parameters that we estimate can also be used to construct safety leaderboards that capture the risk of various agentic actions, which can be used by practitioners for audit prioritization. Applying our methods to tau-bench Retail, ToolSandbox, and ImpossibleBench, we find that intervention effects vary substantially across agent actions and states. More broadly, our work suggests a rich research agenda at the intersection of causal inference and agentic evaluation, which we believe is essential for the reliable and safe deployment of AI agents in the real world.
Read the full abstract
Artificial Intelligence
Assessing the Impact of Retrieval-Augmented Generation on Algorithmic Problem-Solving
Competitive programming problems pose distinctive challenges for Large Language Models, as they demand not only proficient code generation but also advanced algorithmic reasoning. This study presents a Retrieval-Augmented Generation approach specifically designed to enhance the capacity of Large Language Models to address such problems. Our approach integrates an embedding-based retrieval mechanism and a similarity search algorithm to identify analogous problems within a knowledge database, employing a dynamically adjusted threshold to mitigate retrieval noise. We conducted an evaluation using the CodeContests dataset, focusing on the test split problems that exhibit similarity to instances from the training split. The experimental framework considered three conditions: (i) problems provided solely with their original descriptions, (ii) problems supplemented with randomly selected examples from the training split, and (iii) problems augmented with similar examples retrieved through the proposed method. In addition, we incorporated a Min-K membership inference analysis to examine whether the generated code solutions were influenced by potential memorization of training data within the evaluated Large Language Models. This complementary perspective enables a deeper understanding of model behavior by distinguishing genuine generalization through retrieval from implicit recall of pretraining samples. Performance was assessed using both pass rates and the CodeForces Elo rating of the problems. The results provide empirical evidence that similarity-based retrieval substantially enhances Large Language Models performance in competitive programming scenarios, offering an scalable framework for improving code generation in algorithmic problem-solving contexts.
Read the full abstract
Econometrics
Box-Cox Transformations of Correlation Matrices
We consider Box-Cox transformations of correlation matrices, $\Phi_\alpha(C)=\mathrm{vecl}\{(C^{\alpha}-I)/\alpha\}$, with Jacobian $J_\alpha$. The family nests the raw correlations at $\alpha=1$, the generalised Fisher transformation (GFT), $C\mapsto\mathrm{vecl}(\log C)$, at $\alpha=0$, and the negated off-diagonal concentration coordinates at $\alpha=-1$. The inverse problem is a strictly convex program for every $\alpha$, so every member is injective. The map is onto $\mathbb{R}^{d}$, $d=n(n-1)/2$, if and only if $\alpha\le0$, such that $\alpha=0$ defines the boundary for unrestricted parametrizations. For $\alpha>0$ the image is bounded. The conditioning obeys the bound $\kappa(J_\alpha)\le\kappa(C)^{|1-\alpha|}$, where $\kappa(C)=\lambda_{\max}(C)/\lambda_{\min}(C)$, and on the surjective range the exponent is smallest at $\alpha=0$. Under Gaussian sampling with $n\geq3$, the asymptotic covariance of the transformed sample correlations is the identity at independence for every member, but only $\alpha=0$ retains the identity to first order in the correlations, whereas the leading joint third cumulant at independence vanishes only at $\alpha=1/3$. Finally, applying the transformation to the diagonal rescaling of $C$ that makes the result hollow adds a new family to the existing correlation parametrizations. For every $\alpha$ the rescaling solves a strictly convex problem and the inverse is explicit, and the scaled coordinates are unrestricted exactly when $\alpha=0$, where they recover the known global normal coordinates, centred at independence, for the efficient information metric of the correlations.
Read the full abstract
Signal Processing
On PMU Adequacy for Wideband Dynamics Tracking Beyond Attenuation Assumptions
Recent work highlights the inadequacy of standard synchrophasors in capturing high-frequency grid dynamics, including data center oscillations, primarily due to filtering and windowing attenuation during phasor estimation. This letter shows how synchrophasor ambient data spectrum itself contains information on attenuation, which provides a way to identify critically impacted measurements and potentially estimate and undo the effect. Results are shown for synthetic data as well as actual synchrophasor estimates from digital fault recorders (DFRs) within the Dominion Energy system.
Read the full abstract
Nuclear Theory
Possible Solution to the Gallium Anomaly? Nuclear Transition Densities
The gallium anomaly, a 20% deficit in the measured $\nu_e$ captures on $^{71}$Ga, remains unexplained. A recent work [Phys. Rev. Lett. 137, 131805 (2026), arXiv:2512.20560] has proposed a solution based on the breakdown of the factorization of lepton and nuclear wavefunctions that underlies the prediction of the cross section, combined with ad hoc Gamow-Teller nuclear transition densities. We test this proposed solution with transition densities calculated within the nuclear shell model (NSM), normalized to the $^{71}$Ge electron-capture half-life and convolved with Dirac-Hartree-Fock-Slater electron wavefunctions. We explore several NSM interactions, bare and effective Gamow-Teller operators, as well as the impact of the valence space. In all cases, we obtain remarkably consistent NSM transition densities. Our factorized and nonfactorized cross sections agree, remaining 20% higher than the measured value. In addition, we use NSM single-nucleon transition densities as ansatz and find that solving the anomaly requires fine-tuned densities, with percent-level cancellations and an overall scale orders of magnitude larger than the calculated NSM ones. We conclude that the gallium anomaly is not solved when using realistic nuclear transition densities.
Read the full abstract
Group Theory
Braid Groups in Lean : Formalizing Dehornoy's Subword Reversing Algorithm
We formalize the group-theoretic definition of infinite and finite Artin braid groups in Lean. We implement a polynomial-time procedure for determining equivalence in the braid group known as Dehornoy's subword reversing algorithm. Termination and correctness of this algorithm are shown using rectangular grids, which are a specific form of van Kampen diagram. The literature only hints at a proof linking rectangular grids to braid equivalence. We present a novel, three-stage proof of this result, introducing partial grids and associated size metrics.
Read the full abstract
High Energy Physics (Theory)
Edge dynamics of covariant higher-rank Chern-Simons theory of fractons
We study the covariant higher-rank Chern-Simons theory of fractons in 2+1 dimensions in the presence of a planar boundary. The differential generator associated with the scalar gauge symmetry singles out a single scalar Ward combination of the functional field equations. Its zeroth and first normal moments give two integrated boundary identities. The zeroth moment yields a traceless rank-two current obeying a fractonic double-divergence conservation law and a higher-rank Kac-Moody algebra. The first moment makes the normal-tangential field $w_i=h_{2i}$ explicit and yields an ordinarily conserved vector current with a standard $U(1)$ Kac-Moody algebra. Functional differentiation also produces a mixed equal-time bulk-boundary commutator between $w_1$ and $h_{11}$: it is non-vanishing at finite normal distance and vanishes in the intrinsic boundary limit. The equal-time commutators of the boundary current with the spatial components of the gauge-invariant field strength vanish, so microcausality is satisfied; when the boundary parameters are such that the two boundary sectors decouple, its retarded response vanishes at all times. Locally, the boundary constraints are solved by $w_i=\partial_i\chi$ and $h_{ij}=\partial_i\partial_j\xi+\eta_{ij}\phi$. When the boundary parameters are such that the two sectors decouple, the induced action is the direct sum of a chiral boson in the $w_i$ sector and a higher-rank sector in which $\phi$ is auxiliary and $\psi=\partial_1\xi$ is the propagating field. The latter has a non-chiral Luttinger description on its positive-energy branch and admits a chiral reduction. The boundary conditions determine the velocity and positivity branch of the $w_i$ mode, while the compatibility analysis in the $h_{ij}$ sector selects the positive-energy branch. On the positive-energy branch the two chiral modes co-propagate with orientation opposite to the sign of $\kappa$.