Differential Geometry
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Thursday 3 September
The classical technique of moving frames allows to effectively compute differential invariants of group actions. A standard observation is that the expressions of differential invariants contain radicals and more complicated algebraic functions, however the global Lie-Tresse theorem claims that for algebraic pseudogroup actions the algebra of differential invariants is generated by rational functions in jet variables. Here, we explain this apparent contradiction and illustrate it in a few well-elaborated examples. Constructively, our result gives a method to algorithmically compute global differential invariants.
Data Structures and Algorithms
Diffuse Gaussian Truncation For Deterministic Approximate Counting
We give deterministic FPTASes for two dense counting problems on which the known deterministic algorithms, based on zero-free interpolation, run in quasipolynomial time. For fixed $0<\gamma<1/2$ and $0<\theta\leq1$, the first approximates $\mathrm{haf}(A)$ for a symmetric matrix $A$ when its support graph $G$ has minimum degree at least $(1/2+\gamma)n$ and its nonzero entries lie in $[\theta,1]$. It also approximates permanents under the analogous bipartite condition, including full-support matrices in $[\theta,1]$. For fixed $\beta>0$ and $0<\kappa\leq1$, the second approximates the zero-field Ising partition function $Z(J)$ for zero-diagonal real symmetric matrices $J$ satisfying $\max_{i,j}|J_{ij}|\leq\beta/n$ and $\lambda_{\max}(J)\leq1-\kappa$. No separate lower-eigenvalue condition is imposed. We further prove $\log\mathrm{haf}(A)=h_A(G)-n/2+O_{\gamma,\theta}(1)$ and $Z(J)=2^n\det(I-J)^{-1/2}(1+O_{\beta,\kappa}(1/n))$. Here $h_A(G)$ is the maximum weighted fractional-matching entropy. For unweighted graphs, the first formula improves the Cuckler–Kahn error from $o(n)$ to $O_\gamma(1)$ on the fixed-margin class and extends it to weights in $[\theta,1]$. Both algorithms use a common Gaussian truncation principle. Each problem becomes an integral of a product of a fixed entire function over Gaussian coordinates, with possibly indefinite moment matrix entries of order $1/n$. Cancelling the linear term and exactly resumming the quadratic term leaves a coordinate remainder vanishing to order at least three. Complex dilation handles small supports. For large supports, we bound the recombined tail by a large-deviation rate that beats the entropy of the subsets. The truncation error is at most $(CR/n)^{R/2}+e^{-cn}$. This faster-than-geometric decay permits $R\log(en/R)=O(\log n+\log(1/\epsilon))$ and hence polynomial enumeration.
Read the full abstract
Instrumentation and Detectors
Monitoring antiproton numbers with a CMOS detector in a dense-track environment
The production of antihydrogen by the GBAR experiment at AD/ELENA requires good knowledge of the number of incident keV antiprotons, which can be problematic. We have used a commercial CMOS digital camera mounted around the experimental vacuum chamber to determine antiproton numbers from ionising particles created in the annihilation process on the surface of microchannel plate detectors which are used for beam imaging. We show that the multiplicity of emerging charged particles is as expected for individual annihilations of antiprotons with nucleons at rest, taking into account the surrounding material budget. Most of those particles are in the minimal ionising regime, but can be detected with nearly 100% efficiency in the CMOS pixel detector, while due to the thin depletion layer the device is insensitive to background gammas. Thanks to the high granularity and small pixel size millions of antiproton annihilations can be reconstructed in a dense tracking environment over a large dynamic range with good resolution. From cluster length studies of non perpendicular tracks the thickness of the depletion zone and effective detection area was estimated. The cluster length also allows for a monitoring of track angles. Antiproton numbers are determined from the number of reconstructed clusters in the CMOS sensor by means of the covered solid angle relative to a calibration measurements with well known beam intensities at the most upstream location of the GBAR apparatus. Material effects on the emerging annihilation products were estimated by Monte Carlo (Geant4) calculations, while annihilation artefacts on the complex surface of a microchannel plate are cancelled out in this approach. This method minimises largely systematic uncertainties, leading to a final error of roughly 10% for the reconstruction of absolute antiproton numbers.
Read the full abstract
Astrophysics of Galaxies
Absolute Motion of the Infrared Counterpart to Sagittarius A* in the Gaia Celestial Reference Frame 3 and Limits on an Intermediate-mass Black Hole Companion
We report the first proper motion and acceleration measurements of the infrared (IR) counterpart to Sagittarius A* (Sgr A*-IR), the supermassive black hole (SMBH) at the center of our Galaxy, in the Gaia-Celestial Reference Frame (Gaia-CRF3). This reference frame realizes the International Celestial Reference System (ICRS), which is an absolute reference coordinate system defined by quasars. A combination of Gaia and Hubble Space Telescope data was used to transform Keck adaptive optics (AO) observations into Gaia-CRF3. We developed a method for selecting reference stars that minimizes astrometric transformation errors (statistical error = $0.10-0.63$ mas) and drift of the coordinate system (systematic error $\sim 0.01$ mas/yr). We find the proper motion of Sgr A*-IR in Gaia-CRF3 to be $\mu_{\alpha^{*}} = -3.093 \pm 0.085$ mas yr$^{-1}$ and $\mu_{\delta}= -5.62 \pm 0.13$ mas yr$^{-1}$ with the initial position at $t_{0} = 2016.0$ of R.A. = 266.41680848 $\pm$ 0.00000029 deg and DEC = -29.00783947 $\pm$ 0.00000050 deg, which translates to a precision of 1.05 mas in R.A. and 1.79 mas DEC. This is consistent with the astrometric measurements of the radio counterpart to Sgr A* by \citet{Xu_2022}. We also place a $2\sigma$ upper constraint of the acceleration of Sgr A*-IR on the sky at 0.061 mas yr$^{-2}$. This acceleration limit on Sgr A*-IR excludes any intermediate-mass black hole companion with mass $\gtrsim 4\times 10^{4}$ $M_{\odot}$ within a distance of $\sim$0.01 pc, consistent with previous studies. With the release of Gaia Data Release 4, we predict these limits will be improved by at least a factor of two.
Read the full abstract
Quantum Physics
Algebraic Operator Decomposition: A Partitioned Architecture for Noise-Resilient Quantum Computing
We present an operator-decomposition architecture that mathematically maps a global operator into independently executable local operators, reducing the maximum quantum circuit depth at the cost of classical reconstruction and sampling overhead. By framing complex Quantum Circuits around an operator in a vector space that can be algebraically pre-decomposed, AOD complements quantum error correction and error-mitigation approaches by performing algebraic decomposition before quantum execution. Our approach leans in the computer science definition of a Monoid: a design pattern and mathematical concept consisting of a data type, a combining function that is associative, and a safe identity (neutral) element that does not change other values when combined. Simulation wise we define a MapReduce programming model where the addition (+) is the reducer, thus leveraging a naturally stable commutative monoid which carries zero "negative-probability tax" or phase conflicts. Furthermore, we define a Vector Space of Linear Operators over Additive Abelian Groups that benefit from this paradigm, including: Inner Products, Series expansions, Traces and Convolutions. Finally, we present the mathematical foundations and simulation results for this paradigm.
Read the full abstract
Cryptography and Security
PatchBench: Evaluating AI Agents for Vulnerability Patching
AI agents have recently demonstrated strong performance in automated vulnerability patching. However, existing evaluations often validate a patch only by testing whether the provided Proof-of-Concept (PoC) input still triggers a crash. This leaves two key threats to validity: agents may reproduce memorized historical developer patches, or they may generate surface-level fixes that only suppress the reported crash. We study these concerns for C/C++ vulnerability patching. We introduce a patch similarity metric to detect memorized patches. On average, 25% of the agent patches exhibit substantial similarity to historical developer patches, indicating that patch memorization is a real threat to the validity of vulnerability patching evaluations. Meanwhile, agents also frequently exploit benchmark structures to pass patch validation by patching on the crash stack trace to suppress the crash, rather than localizing and fixing the root cause of the vulnerabilities. To handle these issues, we propose PatchBench, a new benchmark for evaluating AI agents on realistic vulnerability patching tasks. PatchBench selects vulnerabilities whose ground-truth fixes lie outside the crash stack and uses vulnerability transplant and code mutations to migrate historical vulnerabilities into new repository contexts, reducing the risks of surface-level fixes and patch memorization. We develop new patch validation methods that thoroughly evaluate both security and semantic correctness of agent patches. Across 11 state-of-the-art agents, including the top three AIxCC agents, the original PoC-only validation inflates the patching task solve rate of agents by 1.83$\times$ on average. Our results reveal key limitations of current patching agents and point to future research directions for more reliable vulnerability repair.
Read the full abstract
Materials Science
Synthesis and Characterization of Compositionally Complex (Gd/Ho/Er/Dy)2Zr2O7 Thin Film Combinatorial Library
High-throughput synthesis and characterization of novel ceramic materials with improved thermomechanical properties and phase stability are needed to accelerate the discovery of next-generation thermal barrier materials. A combinatorial thin film material library of (GdDyHoEr)2Zr2O7 were created via combinatorial magnetron reactive sputtering with rare-earth/zirconium alloy targets. Structural, chemical, and thermal property characterization mapping across the four component composition space was performed and correlated with thermal transport measurements. Steady state thermoreflectance mapping identifies a pronounced minimum in thermal conductivity within the Dy/Gd-rich quadrant. This minimum does not coincide with either the equiatomic composition or the region predicted to exhibit maximum cation size disorder. Instead, it corresponds to the largest experimentally observed lattice parameter, despite deviating from Vegard-like chemical averaging, and is independent of grain size and whole-pattern microstrain. These observations suggest that the way the fluorite lattice accommodates compositional complexity, rather than cation size disorder alone, provides a more informative descriptor of thermal transport. Overall, this work establishes a high-throughput workflow for combinatorial thin-film synthesis and multimodal characterization, enabling the rapid identification of previously inaccessible structure-property relationships in compositionally complex ceramics.
Read the full abstract
Quantum Algebra
A Galois connection between subalgebras and tensor subcategories
Let $\mathcal{B}$ be a braided finite tensor category and let $A$ be a simple commutative algebra in $\mathcal{B}$. We construct an order-reversing Galois connection between subalgebras of $A$ and tensor subcategories of $\mathcal{B}_A$. Let $\mathcal{B}'$ denote the Müger center and set $A':=A\cap\mathcal{B}'$. The closure operators are $B\mapsto\langle B,A'\rangle_{\mathrm{alg}}$ and $\mathcal{E}\mapsto\langle\mathcal{E},\mathcal{B}_A^{\mathrm{loc}}\rangle_{\otimes}$. Thus the closed subalgebras are those containing $A'$, while the closed tensor subcategories are those containing $\mathcal{B}_A^{\mathrm{loc}}$; equivalently, the fixed-point intervals $[A',A]_{\mathrm{alg}}$ and $[\mathcal{B}_A^{\mathrm{loc}},\mathcal{B}_A]_{\otimes}$ are anti-isomorphic as lattices. For a finite tensor category $\mathcal{C}$, the canonical algebra in $\mathcal{Z}(\mathcal{C})$ yields an anti-isomorphism between its subalgebras and tensor subcategories of $\mathcal{C}$. When $\mathcal{B}$ is nondegenerate, Frobenius extensions correspond to unimodular tensor subcategories. For a Hopf algebra in $\mathcal{B}$, the correspondence specializes to an order-preserving bijection between Hopf ideals and normal left coideal subalgebras.
Read the full abstract
Algebraic Geometry
Oort's conjecture for split unitary Shimura varieties
We prove that generically on the basic stratum of split unitary Shimura varieties, the universal abelian variety with endomorphism structure and polarization has minimal automorphism group in a precise sense, except for a few degenerate cases. This is a direct analogue of Oort's conjecture on automorphisms of supersingular abelian varieties. On the way, we explicitly compute the generic automorphism group of the universal $p$-divisible group in a given basic isogeny class.
Read the full abstract
Computer Vision
TAP-Path: Task-Adaptive Structural and Token Pruning for Efficient and Trustworthy Pathology Foundation Models
Pathology foundation models improve transferable representation learning for histopathology, but recent gains often rely on encoders with hundreds of millions of parameters and high inference cost. We propose TAP-Path, a task-adaptive compression framework that directly restructures a pretrained Virchow2 encoder rather than distilling it into a separate student. TAP-Path combines validation-driven transformer-block selection, physical removal of redundant blocks, input-adaptive patch-token pruning, multi-depth feature recovery, and a lightweight gated task head. The final model retains 24 of 32 transformer blocks and 70% of patch tokens after pruning, reducing encoder parameters by 24.96% (631.24M to 473.70M) and analytical encoder compute by 35.20% (340.13G to 220.40G FLOPs). Across three task-head optimization seeds, TAP-Path achieved $87.98 \pm 0.067%$ test accuracy, $81.26 \pm 0.49%$ balanced accuracy, and $82.38 \pm 0.48%$ macro-F1 on a 32-class histopathology benchmark, compared with 86.89% for full Virchow2 and 87.67% for UNI2-h. TAP-Path achieved a Brier score of $0.1800 \pm 0.0005$ and failure-detection AUROC of $0.9047 \pm 0.0060$. A validation-only rare-aware objective improved rare-class balanced accuracy in a secondary operating analysis. Frozen external evaluation on 433 CPTAC samples yielded $91.22 \pm 0.83%$ accuracy and $91.10 \pm 0.81%$ balanced accuracy. These results show that task-adaptive structural and token sparsification can improve the accuracy-efficiency trade-off of large pathology foundation models while preserving reliability under internal and external evaluation.
Read the full abstract
Computer Vision
Continuous Actions from Discrete Minds: Latent-Aligned Planning for End-to-End Autonomous Driving
Bridging the gap between the discrete reasoning of Vision-Language Models and the continuous, physics-constrained nature of autonomous driving remains a significant challenge. In this work, we introduce LaPla, a unified Vision-Language-Action (VLA) framework featuring latent-aligned planning to seamlessly ground semantic understanding in precise motion execution. We first design an action tokenizer based on a residual vector-quantized variational autoencoder (VQ-VAE), capturing vehicle kinematics and encoding trajectory features into a structured latent space. Rather than discrete codebook lookups that inevitably introduce quantization errors, LaPla repurposes this representation as a physical prior to bridge the modality gap between high-dimensional semantics and the raw action space. Specifically, given multimodal inputs integrating multi-view images, historical actions, and textual instructions, LaPla incorporates concurrent action queries to causally attend to the multimodal context in a single forward pass, projecting hidden states directly into the pretrained VQ-VAE latent space. The frozen decoder then translates these continuous latents into actions, effectively eliminating quantization errors and ensuring physically plausible trajectories while bypassing time-consuming autoregressive generation. Extensive experiments on the nuScenes benchmark demonstrate that LaPla achieves competitive open-loop performance, reducing long-horizon L2 error by 15.52% compared to state-of-the-art VLA methods. Closed-loop evaluations on the NVIDIA AlpaSim simulator further confirm its superior capability in ensuring smooth driving progress, improving the success rate by 33.34 percentage points with significantly reduced inference latency.
Read the full abstract
Combinatorics
Positive and Negative Square Energies of $2$-Connected Graphs
Let $G$ be a graph of order $n$, and let $s^+(G)$ and $s^-(G)$ denote the sums of the squares of the positive and negative adjacency eigenvalues of $G$, respectively. Recently, Liu, Tang, and Zhang proved the conjecture of Elphick, Farber, Goldberg, and Wocjan that every connected graph $G$ of order $n$ satisfies $ \min\{s^+(G), s^-(G)\} \ge n-1. $ For positive square energy, we strengthen this result by showing that every $2$-connected graph $G$ of order $n$ which is not a cycle satisfies $s^+(G)\ge n$. The formally analogous assertion for $s^-$ is false: the complete graph $K_n$ satisfies $s^-(K_n)=n-1$. We prove a natural counterpart in the triangle-free class: every triangle-free $2$-connected noncycle $G$ satisfies $ \min\{s^+(G),s^-(G)\}>n. $ More generally, it is enough that some maximum-degree vertex of $G$ belongs to no triangle. Together with the exact square energies of cycles, this characterizes the triangle-free $2$-connected graphs for which $s^-(G)\ge n$; the only exceptions are the cycles $C_{4k+3}$ with $k\geq1$.
Read the full abstract
Theoretical Economics
Tournaments with Managerial Discretion
We study tournaments with managerial discretion in hiring. A manager selects a coworker from a pool of candidates and then competes against him in a Lazear–Rosen–style tournament with a prize equal to a share of total output. A profit-maximizing principal sets the prize share together with a head start (or handicap)—an advantage (or disadvantage) in the output comparison—granted to the manager. The head start affects output through three channels: (i) encouraging the manager, (ii) discouraging the new hire, and (iii) inducing the manager to hire a stronger candidate. The hiring effect dominates the discouragement effect until the strongest candidate is hired; beyond that point, any further head start discourages the new hire more than it encourages the manager. The optimal contract therefore grants a head start just large enough to induce the manager to hire the strongest candidate.
Read the full abstract
Numerical Analysis
A Multigrid Method for CutFEM and its Convergence
We develop a convergence theory for geometric multigrid with vertex-patch smoothers applied to cut finite element discretizations of the Poisson problem. The framework addresses non-inherited level forms and the mismatch between the physical and active domains. Using the discrete extension property, we prove two-level convergence bounds uniform in the mesh size and the cut geometry, and W-cycle bounds under an additional smallness assumption on the two-level rate. The numerical experiments intentionally use the stronger V-cycle, for which no convergence bound is claimed here. The convergence constants degrade with the degree $p$. Lowering the ghost penalty improves iteration counts. An aligned two-cell model exhibits a semidefiniteness threshold of order $p^{-2}$, whereas the visibility scale of a degree-$p$ cut mode decreases exponentially. Experiments at the model threshold reduce the iteration counts, but do not establish an assembled-operator threshold.
Read the full abstract
Machine Learning
Subspace Inference Enables Efficient Active Reward Learning from Preferences
Reinforcement learning from human feedback (RLHF) has emerged as a powerful yet sample-inefficient approach for learning reward models from human preferences, making active learning a critical component in synthesizing informative preference queries. However, effective uncertainty quantification required for active learning remains a key challenge for large neural network reward models. In this paper, we introduce PreferenceEKF, a sample-efficient approach that tracks reward model uncertainty by framing active preference learning as a sequential Bayesian filtering problem. Instead of relying on computationally prohibitive posterior inference over the full neural network parameter space, our method performs sequential inference via an extended Kalman filter within a low-dimensional parameter subspace, continuously updating the reward model posterior as new preference queries arrive. Our approach enables scalable sampling of neural network parameters to efficiently compute acquisition functions for active reward learning. Experiments on the D4RL and V-D4RL benchmarks demonstrate that our approach achieves better sample efficiency, runtime, scalability, and calibration compared to other Bayesian deep learning approaches, and the learned reward models lead to competitive offline reinforcement learning policy performance. This highlights the potential of scalable Bayesian methods for preference-based reward modeling in RLHF. Our code is available at https://github.com/yutaizhou/bnn_pref.
Read the full abstract
Quantum Physics
Small-quench Loschmidt dynamics near quantum critical points
We study the short-time Loschmidt dynamics after a small sudden quench near a quantum critical point. We show that the initial quadratic growth of the Loschmidt rate function is governed by the variance of the quench operator per system size. For a local quench operator, this variance density is exactly equal to the spatial sum of the equal-time connected two-point correlation function in the initial ground state. This relation connects the early-time Loschmidt response to static critical correlations. Assuming power-law correlations at criticality, we classify the finite-size scaling of the short-time coefficient by the scaling dimension of the quench operator. In one dimension, the coefficient is finite, logarithmically enhanced, or algebraically enhanced with system size. We illustrate this operator dependence in the transverse-field Ising chain, where transverse-field and longitudinal-field quenches couple to different critical operators. We also discuss the first correction beyond the quadratic regime using the fourth cumulant of the post-quench Hamiltonian.
Read the full abstract
Quantum Physics
Quantum-Enhanced Phase Estimation with Photon-Added Even and Odd Coherent States in an SU(1,1) Interferometer
We investigate phase estimation in an SU(1,1) interferometer employing $m$-photon-added even and odd coherent states as nonclassical input resources. The phase sensitivity is evaluated through intensity detection and the error propagation method, while the ultimate precision limit is determined from the quantum Cramér-Rao bound with the quantum Fisher information serving as the relevant metrological quantity. Our results demonstrate that photon addition significantly enhances the phase sensitivity, increases the quantum Fisher information, and reduces the quantum Cramér-Rao bound, leading to a clear improvement over the corresponding even and odd coherent states. Furthermore, the achievable sensitivity exceeds the standard quantum limit and gradually approaches the Heisenberg scaling with increasing photon-addition number. We also find that the $m$-photon-added even coherent states exhibit a modest advantage over their odd counterparts. As $m$ increases, however, this distinction becomes progressively weaker, suggesting that photon addition diminishes the role of the initial parity of the coherent state in determining the interferometric performance.
Read the full abstract
Artificial Intelligence
Spurious Advantage Hidden in GRPO
Group Relative Policy Optimization (GRPO) is widely studied for reinforcement learning with verifiable rewards, where its advantage estimator assigns each rollout a magnitude from within-group reward statistics. In the common case, this magnitude rewards rollouts that reach the correct answer through reasoning. Yet, an overlooked case shares the same surface: a rollout may land on it by guessing, and the formula still assigns a high magnitude, which we identify as the spurious advantage. This arises in three cases: bounded-answer tasks with a small candidate set; open-answer sets hosting bounded sub-cases; and search agents whose budget opens many paths to the same answer. In all three, this misleads the policy toward guess-like behaviors. We propose SIGNBALANCE, whose magnitude is composition-free: it keeps the verifier sign, uses a global scale, and restores zero-mean balance via a stop-gradient per-class rescaling. Across math and search agent benchmarks at different scales, SIGNBALANCE matches GRPO on open-answer math and improves on bounded-answer math and search agents. Code will be released.
Read the full abstract
Soft Condensed Matter
Topological Mixing and Braiding Universality in Polar Active Matter
Connecting the autonomous kinematics of active matter to its emergent macroscopic transport is constrained by the requirement for high-resolution Eulerian velocity fields. Here, we use confined bacterial suspensions as a model active fluid, mapping the sparse Lagrangian trajectories of fluorescent spy cells into (2+1)-dimensional geometric braids that directly encode the spatiotemporal entanglement of the flow. We use the finite-time braiding exponent (FTBE) as a proxy to quantify the topological entropy and chaotic mixing of confined bacterial suspensions. In moderately confined wet systems, we find that hydrodynamic coupling drives a structural transition from a dilute active gas to coherent vortices, and ultimately to active turbulence, revealing three distinct density-dependent regimes of topological mixing. Conversely, truncating the hydrodynamic screening length via extreme confinement drives the system toward a dry active matter limit. In this limit, dense in-plane steric collisions suppress irreducible entanglement and substantially reduce the FTBE at high particle densities. By evaluating the topological complexity generated per encounter, we reveal a transition from a discrete geometric encounter regime to an areal escape mechanism. Finally, we establish a square-root scaling between the FTBE and the effective diffusivity, placing the self-sustained mixing of active fluids into the pathline braiding universality class.
Read the full abstract
Software Engineering
When Models Edit Too Much: On the Fidelity of Minimal Code Edits
Large language models (LLMs) are increasingly used to edit existing code, but correctness alone is not enough: useful repairs should also be minimal, reviewable, and faithful to the original implementation. We study over-editing, the tendency of a model to rewrite code beyond what is required to fix a bug. We construct an evaluation framework from 400 BigCodeBench problems by injecting controlled AST-level corruptions into reference solutions, giving each repair task a known minimal patch. Across frontier LLMs, over-editing is widespread even among strong models like GPT-5.5: high Pass@1 can coexist with unnecessarily large edits and added cognitive complexity. A preservation instruction substantially reduces this behavior, lowering average excess Levenshtein distance from 0.195 to 0.131, reducing added cognitive complexity by 26.6%, and increasing Pass@1 by 2.3 points. However, these gains do not simply follow from a larger reasoning budget or larger models. We next ask whether minimal editing can be learned directly during post-training. We observe that supervised fine-tuning overfits to seen corruption patterns, whereas reinforcement learning gives the best out-of-domain edit-fidelity and performance-retention trade-off. These results position edit fidelity as a distinct axis of code-repair quality and show that it can be measured and learned.
Read the full abstract
Mathematical Physics
A simple derivation of the zero-temperature BCS functional from the reduced BCS Hamiltonian
We consider a grand-canonical reduced BCS Hamiltonian on the full fermionic Fock space with an attractive pairing interaction supported in an energy shell around the Fermi surface. For fixed chemical potential, we prove that the exact finite-volume ground-state energy differs from the minimum of the corresponding finite-volume BCS functional by a volume-independent error. As a consequence, the BCS variational principle is exact for the ground-state energy density in the thermodynamic limit.
Read the full abstract
Data Structures and Algorithms
Batched Pandora's Box
Motivated by numerous parallelizable stochastic search problems, most notable and timely among them being LLM inference-time scaling, we propose and study batched versions of the Pandora's Box problem of Weitzman. In particular, boxes are opened in capacity-constrained batches, each batch has a setup cost, and all rewards in a batch are revealed together. We consider two different variants, motivated by different application environments: one where boxes are reusable (i.e., can provide multiple i.i.d.~samples) and another where they are not. For both variants we rule out most “simple” natural heuristics, and also formally prove NP-hardness of approximation in the traditional sense. We then relax the problem to allow bi-criteria approximations, with respect to both rewards and setup costs, where we exhibit constant approximation algorithms for both the reusable and non-reusable settings. This is obtained through a linear-programming relaxation of Pandora's Box problem, followed by randomized or Pipage rounding.
Read the full abstract
Cryptography and Security
AI-Assisted Design of a Post-Quantum Cryptographic Accelerator: A Deployed-Silicon Case Study
Post-quantum migration is mandated on published timelines, and silicon that ships with a defect cannot be patched remotely. The standard acceptance gate cannot detect an entire class of ML-DSA defects. Signing resamples until a candidate meets its norm bounds, so the executed path varies with the message, whereas known-answer tests (KATs) sample fixed values and reach only the depths their seeds trigger. Our accelerator passed its full KAT regression while carrying a norm check that outran block-RAM latency, leaving each candidate's final coefficients unverified; the escape surfaced at reject-loop iteration 5. The blind spot lies in the instrument, not the engineer; care cannot remove it. We replace that gate. A byte-exact golden-reference oracle paired with randomized adversarial soak drives the rejection loop past any fixed vector, closing the gap: 301,343 data-dependent signings, zero escapes. Because the gate judges artifacts and never authors, trust becomes separable from authorship, making AI authorship an answerable question. We report 232 logged experiments in which an agentic large language model drove a unified ML-KEM-768 and ML-DSA-65 accelerator with on-chip key custody from RTL to PCIe bring-up on one Kintex-7 XC7K160T, shipped at 98.5% slice occupancy. Success was 71.6%, following a hardware-coupling gradient, 77-85% for documentation and research against 50-53% for synthesis and bring-up, which observability can explain: failure concentrates where corrective signals are physical-side only. That so unreliable an author produced an artifact byte-exact across all six FIPS operations – its deployed baseline surviving the same 779,945-check zero-failure soak – is the claim.
Read the full abstract
Mathematical Physics
Limits of Stochastic Semigroups and Block-Triangular Majorisation
We investigate limits of semigroups of stochastic matrices defined by their invariant distribution. Given probability vectors $q(\beta)$ depending on a parameter $\beta$, we introduce a notion of convergence as $\beta\to\infty$ for the corresponding semigroups of $q(\beta)$-preserving stochastic matrices and investigate the structure of the resulting limit. In general, the limiting semigroup differs from the semigroup preserving the limiting distribution, showing that these two operations do not commute. We develop a general framework for such limiting semigroups and study in detail the case in which the invariant distributions are Gibbs vectors at inverse temperature $\beta$. In that case, the limiting semigroup consists of block-upper-triangular stochastic matrices subject to additional substochasticity constraints. We characterise and enumerate their extremal elements and determine the preorder on probability vectors induced by the action of the semigroup. The resulting notion of Block-Triangular majorisation interpolates between ordinary majorisation and upper triangular (also known as unordered) majorisation. We show that it is completely characterised by a finite family of monotones and analyse the corresponding behaviour of Rényi $\alpha$-entropies as $\beta\to\infty$.
Read the full abstract
Optimization and Control
Factor-parity Hall sets and controllability
We introduce a class of Hall sets, which we call factor-parity Hall sets, whose elements split into good and bad brackets. We prove that the good brackets can be steered simultaneously and arbitrarily in small time, which yields sufficient conditions for the small-time local controllability of control-affine systems. This positive result draws on constructions of Kawski, Agrachev-Gamkrelidze and Krastanov. Conversely, we prove that each bad bracket generates an obstruction to controllability, hence a family of necessary conditions.
Read the full abstract
Software Engineering
LabelMate: An LLM-Driven Framework for Refined Issue Report Labeling
Software users often submit issue reports to a product's issue tracking system to report defects, suggest enhancements, or raise other product-related concerns. Labeling these issue reports supports effective planning and improves community engagement. However, many issue reports remain unlabeled due to the substantial manual effort required to design an appropriate label taxonomy, then assign suitable labels from this taxonomy to new issue reports. Existing automated labeling approaches attempt to mitigate these challenges. However, they suffer from key limitations, such as extensive manual intervention, the assignment of generic labels, and a dependence on existing labeled datasets. To address these limitations, we propose LabelMate, a novel Large Language Model (LLM)-driven framework that (1) derives a comprehensive, project-specific label set from historical issue reports and (2) automatically assigns relevant labels to new issue reports without requiring any pre-labeled training data. We evaluate LabelMate on 16,500 issue reports from 30 popular and diverse GitHub repositories. Based on this dataset, our approach generates a coherent list of 275 labels and achieves an average labeling accuracy of 89.84%, a statistically significant improvement over existing generic label assigning approaches. These results demonstrate that LabelMate offers an efficient, domain-adaptive solution to streamline the issue labeling process.
Read the full abstract
Algebraic Geometry
An obstruction to the maximal singularity of the Hilbert scheme of points on threefolds
We prove a necessary condition conjecture ([27, Conjecture B]) for the maximal singularity of the Hilbert scheme of points in 3D for a large class of degrees, which in particular implies the 1978 Briançon-Iarrobino Conjecture for a tetrahedral degree, and vastly generalises the result in [19]. The novelty lies in introducing a new upper bound on the dimensions of the tangent spaces of the claimed non-maximally singular ideals, constructing canonical ideals that satisfy the claimed maximal singularity condition, providing a lower bound on the dimension of their tangent spaces, and finally comparing the two cases.
Read the full abstract
General Relativity and Quantum Cosmology
An off-shell conformally invariant Galilean Weyl tensor
We propose a manifestly conformally invariant off-shell definition of the Weyl tensor in Galilean geometry. We also propose definitions for the electric and magnetic parts thereof. For the latter to vanish, observers with specific kinematical properties need to exist. While this is guaranteed by the Newton–Cartan equation, it might not be true for other Galilean invariant theories. Therefore, we propose that imposing the existence of observers for which the off-shell magnetic part is zero should be a necessary condition for a Galilean invariant theory to be called `Newtonian', in the spirit of the `Newtonian' condition, introduced by Trautman, in standard Newton–Cartan gravity. As a side result, we also show that there exists a unique Galilean boost-invariant connection that can be built from a Galilean structure and a choice of Coriolis field, even when the clock form is not closed. No extra structure is needed, contrasting with the standard approach used to construct boost-invariant connections which introduces a mass gauge field.
Read the full abstract
Analysis of PDEs
Caffarelli Estimates under Lipschitz Perturbations
We study dimension-free differential estimates for Brenier maps under first-order perturbations of the two marginals. Caffarelli's contraction theorem yields such estimates when the source and target potentials satisfy a pointwise Hessian comparison. We show that the same conclusion remains true under arbitrary globally Lipschitz perturbations of both marginals, without additional assumptions. In particular, this answers the conjecture due to Fathi, Mikulincer, and Shenfeld. More precisely, let $d\ge1$, $L\ge0$, and let $B:\mathbb{R}^d\to\mathbb{R}$ be globally $L$-Lipschitz. We prove that the Brenier map from the standard Gaussian measure $\gamma_d$ to the probability measure proportional to $\mathrm{e}^{-B}\gamma_d$ has a globally Lipschitz representative whose bound depends only on $L$. The Brenier potential also belongs to $C^{1,1}(\mathbb{R}^d)$ and satisfies dimension-free two-sided Hessian bounds. Let $\mathfrak{C}(0,L)$ denote the upper-bound constant, then \[ \log\mathfrak{C}(0,L)=4L^2+\log L+\mathcal{O}(1) \qquad(L\to\infty), \] and a one-dimensional example with Brenier map $T_L$ satisfies $\log\operatorname{Lip}(T_L)\ge L^2/2$. We deduce this Gaussian estimate from an anisotropic two-marginal result. Let $V,W:\mathbb{R}^d\to\mathbb{R}$ be potentials, and let $Q,P$ be positive-definite matrices. Using the distributional curvature bounds $D^2V\preceq Q$ and $D^2W\succeq P$, we establish matrix Hessian bounds for Brenier maps between arbitrary globally Lipschitz perturbations of the reference marginals. Most importantly, the constants keep the directional geometry of the essential gradient ranges. For affine perturbations, our estimates recover the sharp noncommuting Caffarelli tensor. We further obtain pointwise displacement bounds for Gaussian perturbations and spectral Hessian estimates for compact Gaussian mixtures.
Read the full abstract
High Energy Astrophysical Phenomena
Sgr A* as a Galactic PeVatron: Multimessenger Signatures of the Magnetic Penrose Process
The magnetic Penrose process (MPP) is one of the most efficient mechanisms for extracting rotational energy from a magnetized Kerr black hole (BH), enabling charged particles to reach very-high energies. We investigate this process in Sagittarius A* (Sgr A*), the supermassive BH at the Galactic center (GC), focusing on its ultra-high-efficiency regime, in which neutrons undergo beta decay inside the BH ergosphere. As a key novelty, we compute the neutron production spectrum in the accretion flow directly from the nuclear reaction kinematics and follow neutron trajectories in the Kerr spacetime to determine the population that reaches the ergosphere and undergoes the MPP. From this population, we derive the spectrum of accelerated protons and show that Sgr A* can accelerate them up to PeV energies, strengthening its interpretation as a candidate Galactic PeVatron. We further compute the gamma-ray and neutrino emission produced through hadronic interactions of the escaping protons in the Central Molecular Zone. The predicted gamma-ray fluxes exhibit distinctive spectral features that could provide an observational signature of the MPP and may constitute a non-negligible contribution to the very-high-energy emission detected by H.E.S.S. and HAWC. The associated neutrino fluxes remain below the diffuse Galactic component inferred by IceCube, but may still contribute to the high-energy emission from the GC. Remarkably, the predicted signals lie within the projected sensitivity of SWGO for all scenarios considered here and, for some models, only a factor of a few below the nominal CTAO sensitivity, while KM3NeT/ARCA and IceCube-Gen2 will provide complementary tests. Our results establish the MPP as an observable mechanism for extracting BH rotational energy, providing a direct connection between horizon-scale physics and multimessenger observations that can be extended to other magnetized BHs.