Operator Algebras
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Tuesday 1 September
We introduce the notion of proper proximality for countable measured groupoids, extending the corresponding notion for countable groups. We prove that this groupoid property is equivalent to a strengthened form of relative proper proximality for the associated groupoid von Neumann algebra. We investigate permanence properties and show, in particular, that proper proximality is preserved under finite direct products, formation of transformation groupoids, and measure equivalence. We also identify broad classes of properly proximal groupoids, including transverse measured groupoids and free product groupoids, and prove that inner amenable groupoids are never properly proximal.
Computer Vision
MeRoPE: Metric Rotary Position Embedding for Camera-Controlled Video Generation
In camera-controlled video generation, geometry-aware positional encodings condition tokens on camera extrinsics and per-token viewing rays. Existing schemes, however, have a scale-dependent failure mode on real-world metric camera trajectories: homogeneous projective encodings cause attention logits and feature norms to grow unbounded with physical translation baselines. We propose MeRoPE (Metric Rotary Position Embedding), a norm-preserving relative camera encoding for attention. MeRoPE encodes relative orientations between calibrated viewing rays with orthogonal rotation blocks, maps raw metric displacements into multi-frequency rotary phases, and adds a disparity-anchored correspondence prior along the epipolar arc. This design strictly preserves feature norms, bounds pre-softmax attention logits regardless of the physical translation scale, and maintains exact invariance to global rigid coordinate changes. Across nuScenes and PanShot, which cover large-baseline trajectories and diverse camera optics, respectively, MeRoPE achieves stronger camera control than prior encodings, with the best consistency between generated camera motion and conditioning poses in both rotation and translation. Code will be made publicly available.
Read the full abstract
Fluid Dynamics
On the Non-Negativity of Longitudinal Velocity Autocorrelations in Homogeneous Isotropic Turbulence: Kinematic Constraints, Convexity, and Spectral Broadening
A fundamental question in the statistical theory of homogeneous isotropic turbulence is whether the longitudinal velocity autocorrelation $f(r)$ remains non-negative and if it can be shown from first principles such as kinematic realizability. While incompressibility forces the transverse correlation $g(r)$ to exhibit a negative loop ($\int_0^\infty r g(r)\,dr = 0$), the sign behavior of $f(r)$ is subtle. We show that positive-definiteness via Bochner's and Schoenberg's theorems structurally fails to enforce pointwise positivity because the 3D spherical projection kernels are sign-changing. By mapping the problem to the one-dimensional spectrum $E_{11}(k_1)$ via the cosine transform and applying Pólya's criterion, convexity of $E_{11}(k_1)$ provides a rigorous sufficient condition for $f(r) \ge 0$. While pure Kolmogorov $-5/3$ scaling is strictly convex, physically realizable spectra with Batchelor ($k^4$) or Saffman ($k^2$) infrared scaling are necessarily concave near $k_1 = 0$. For mature broadband spectra such as the von Kármán–Pao model, high-precision quadrature confirms that the convex inertial bulk dominates, yielding strictly positive tails and demonstrating that apparent negative dips are discrete Fourier artifacts. Also, we construct a smooth, divergence-free, finite-energy narrowband initial field that produces a genuine negative loop ($\min_r f(r) \approx -0.083$), serving as a counterexample to universal non-negativity. Spectral bandwidth is shown as the governing physical parameter, with nonlinear triad interactions rapidly broadening narrowband fields over an eddy turnover time, systematically suppressing negative excursions and preserving $f(r) \ge 0$ in mature turbulence.
Read the full abstract
Analysis of PDEs
Global regularity for 2D gravity water waves with two retreating point vortices
We prove global well-posedness for the two-dimensional infinite-depth gravity water wave system coupled to a pair of point vortices of opposite strengths. The vortex dipole is initially placed deep below the free surface and oriented so that its leading motion is away from the interface. We show that the vortices retreat almost linearly, and that the velocity induced on the free boundary is therefore integrable in time. Combining this decay with Wu's cubic formulation, together with a transition-of-derivatives and localization argument, we obtain global bounds for small localized perturbations of the flat surface.
Read the full abstract
Computer Vision
One Prompt Is Enough: Watermark Laundering Through Foundation Image Models
Invisible watermarks are typically evaluated against predefined perturbations such as compression, blur, noise, cropping, and denoising. Public foundation image models expose a distinct threat: an attacker can submit a watermarked image with a single reconstruction prompt and obtain a visually faithful output from which the invisible watermark can no longer be decoded reliably. We formalize this failure mode as watermark laundering and evaluate it using a joint payload-fidelity profile that combines bit error rate (BER) with visual and semantic preservation. Across six OpenAI and Google image editing models, three representative watermarking schemes, and 1,800 reconstructed outputs, we identify two complementary laundering regimes: OpenAI models produce the strongest payload disruption across the evaluated schemes, whereas Nano Banana 2 shows that DwtDct remains vulnerable under high-fidelity reconstruction. Prompt ablations show that no single removal-oriented instruction is necessary for payload disruption, indicating that the effect is primarily induced by the reconstruction pathway rather than by explicit attack wording. Comparisons with conventional attacks further show that prompt-conditioned reconstruction constitutes a distinct operational attack interface. These findings motivate foundation-model reconstruction as a missing robustness condition in invisible watermark evaluation.
Read the full abstract
Quantum Physics
A Backend-Agnostic MWIS Kernel for Stochastic Unit Commitment with Neutral-Atom Hardware Validation
Quantum hardware is beginning to address structured combinatorial optimisation, but two steps still block practical use: mapping real operational models onto hardware-compatible instances, and converting noisy hardware output back into feasible decisions. Here we introduce a backend-agnostic computational interface that compiles the discrete decision layer of stochastic unit commitment into a move-based maximum-weight independent set (MWIS) problem, while retaining continuous dispatch and feasibility recovery in the classical computational layer. We validate the approach in a green hydrogen scheduling setting and deploy it on the QuEra Aquila neutral-atom quantum processor. This is the first end-to-end industrial scheduling workflow that connects real operational decisions to programmable neutral-atom hardware through a solver-agnostic MWIS representation. Across a 15-day hardware campaign on 50-node instances, hardware-generated solutions after classical refinement match or exceed the dispatch margins obtained from exact MWIS on every day. When scaling to 144 nodes, encoding quality remains stable, while the probability that the full atom array survives, rather than graph embedding, emerges as the dominant bottleneck to further scaling. Together, these results establish a hardware-compatible computational pathway toward larger problem scales, and lay the groundwork for exploring regimes in which exact classical optimisation may no longer scale efficiently.
Read the full abstract
Computational Physics
A Non-Isothermal Viscoplastic Constitutive Model for Clay Slip Surfaces
Clayey slip surfaces control the reactivation and long-term deformation of slow-moving landslides and may experience thermal fluctuations from climate, seasonal ground-temperature changes, or subsurface heat sources. Experiments show that residual shear strength depends on temperature and shearing rate, yet most numerical approaches use temperature-independent strength parameters. We present a non-isothermal viscoplastic constitutive model for clayey slip surfaces implemented with zero-thickness interface elements. It includes temperature-dependent normal and tangential stiffness, progressive degradation of cohesion and friction angle, and rate-dependent viscoplastic slip governed by a non-associated flow rule. Coupling with hydraulic and thermal balance equations allows the interface response to evolve with stress state, temperature, aperture, and accumulated irreversible displacement. Validation against temperature-controlled drained ring-shear tests on bentonite and smectite-rich soils covers heating–cooling, cooling–heating, and combined thermal paths. The simulations reproduce thermal strengthening at slow shearing rates and thermal weakening or limited sensitivity at higher rates. Application to the Congress Street cut benchmark shows that zero-thickness elements improve the representation of strain localization and progressive failure. Increasing temperature progressively degrades interface strength, increases displacement, joint aperture, and shear strain, and accelerates sliding. Temperature-dependent interface degradation can therefore reduce the apparent stability margin of clayey slopes and should be included in slope-stability assessments involving thermal fluctuations.
Read the full abstract
Computation and Language
Ready to Speak: Aligning LLMs for TTS-Friendly Text Generation
Current Large Language Models (LLMs) are primarily optimized for written text, often producing outputs that are grammatically correct and helpful yet poorly suited for spoken delivery via Text-to-Speech (TTS). In this work, we study how to make LLMs natively generate TTS-friendly text, which we frame as a preference alignment problem: instead of relying on downstream rewriting modules, we directly align LLMs to generate text optimized for spoken delivery. We introduce two preference datasets spanning different target domains, CORA and Recipe, which contain paired TTS-friendly and TTS-unfriendly responses. We further propose an evaluation suite combining a pattern-based heuristic metric, a TTS$\to$ASR evaluation pipeline, and a MUSHRA listening study with human judges. Our experiments compare the recently proposed Feature-aware Sampling and Tuning (FaST) framework – leveraging interpretable features instead of a black-box reward model – against an array of alignment baselines on the TTS-friendly generation task. Notably, we found that FaST achieves the best overall tradeoff between TTS-friendliness and helpfulness across various settings. We also identified a strong correlation between our different metrics, highlighting the ability to reliably assess TTS-friendliness via an efficient heuristic.
Read the full abstract
Machine Learning
Explore More, Drift Less: Outcome-Only Reinforcement Learning Can Suffice for Long-Horizon Interactive Agents
Reinforcement learning is a natural way to post-train LLM agents for long-horizon interactive tasks judged only by end-of-task verification, yet a shared belief holds that outcome-only RL soon hits a ceiling on small open models. Recent work therefore compensates around the training with denser rewards, SFT priors, skill libraries, curated memory, or multi-agent orchestration. We argue the ceiling is an artifact of two failures of common practice. Signal starvation: group-relative RL with sparse outcome-only rewards yields a gradient only when a task's rollout group mixes successes and failures, so under-scaled exploration silences exactly the hardest, most instructive tasks. Policy drift: squeezing many updates out of a small task pool degrades the policy itself, as an unanchored objective lets the sampling distribution collapse exactly when saturation has already made informative groups rare. We present CANOPY (Coverage-ANchored On-PolicY RL), a minimalist protocol attacking both directly: scale same-task exploration until the natural signal reappears, keep every update on-policy, KL-anchored, and confined to the agent's own action tokens, then cash in an enlarged interaction budget at test time. On AppWorld, a long-horizon interactive coding benchmark, a Qwen3-14B policy trained with CANOPY through environment interaction alone–without task-specific supervision, auxiliary credit signals, or elaborate agent scaffolding–topped the public leaderboard (Feb. 2026; Test-Normal TGC 86.9, Test-Challenge 67.6), and the same design principles lift Qwen3.5-9B on SWE-bench Verified by 16.6 points. Agentic RL alone thus internalizes long-horizon capability directly into a small open model; we plan to release the complete training stack at https://github.com/AlibabaResearch/SignalCoverageRL.
Read the full abstract
Machine Learning
Post-Training Science for Supervised Fine-Tuning
Every supervised fine-tuning run forces the same chain of decisions, such as learning rate, batch size, LoRA or full fine-tuning, how many epochs, which optimiser, and what data to feed the model. Each of these is typically rediscovered from scratch for every new model and dataset. Here we measure them under one instrument: a sweep that varies one lever at a time, and spans dense and mixture-of-experts models in two families (Qwen3 and Llama), on four real-world customer SFT datasets, for both LoRA and full fine-tuning. These datasets give a controlled testbed: each task carries an evaluation built with the customer, and its training data is produced by iterative supervised fine-tuning that refines model outputs until they pass that evaluation, so the supervised target is internally consistent and the task judge we report against is the criterion the data was built to satisfy. We ask how the optimal learning rate and batch size move with model scale, family, and data, and whether one selection rule transfers across them; what LoRA trades against full fine-tuning, and how its rank and alpha set what the adapter can learn; whether validation loss (or other metrics, such as loss landscape flatness) faithfully ranks downstream quality; whether post-training gains scale with model size and data volume, on a model ladder extended through mixtures-of-experts to 235B parameters; how many epochs to train before general instruction-following erodes; and whether a geometry-aware optimiser improves on AdamW. Each recommendation is paired with a measure of its uncertainty.
Read the full abstract
Mesoscale and Nanoscale Physics
Multiple chiral Majorana states in proximitized magnetic topological insulator heterostructures
Achieving robust topological superconductivity with multiple Majorana channels is a key step for scalable topological quantum computing. To this end, we investigate magnetically doped three-dimensional topological insulator heterostructures proximitized by an s-wave superconductor within a fully three-dimensional extended Bernevig-Hughes-Zhang framework that explicitly accounts for vertical confinement. We show that magnetic exchange coupling, orbital mixing, and bulk band inversion cooperate to generate effective equal-spin p-wave pairing channels supporting multiple chiral Majorana modes. The number of Majorana channels is determined by the confined modes in the vertical direction and the high-Chern-number phases of the normal state, which allows us to derive an analytical criterion for the emergence of the multiple-Majorana topological superconducting phases. The chiral Majorana modes remain robust against strong disorder and moderate symmetry-breaking perturbations. We also demonstrate that a superconducting phase difference defining a vertical Josephson junction acts as a tunable parameter that controls the hybridization, minigap, and effective multiplicity of the low energy Majorana channels. Our results establish magnetic three-dimensional topological- insulator heterostructures as a promising platform for engineering multiple chiral Majorana modes.
Read the full abstract
Atomic Physics
Polarization measurements of the $2s$–$2p_{3/2}$ VUV transition in N$^{4+}$ excited by electron collisions
We present the linear polarization of the $2s$–$2p_{3/2}$ transition in Li-like N$^{4+}$ excited by electron collisions, measured at electron energies between 85 and 1000~eV. The measured polarizations are compared with theoretical values calculated using the Flexible Atomic Code (FAC). Although the overall trend that the degree of polarization decreases with increasing electron energy is reproduced, the theoretical values are systematically larger in magnitude than the experimental values. After examining possible experimental sources of depolarization, we suggest that the discrepancy likely originates from an overestimation in the theoretical model. Further theoretical and experimental investigations along the isoelectronic sequence are needed to resolve this discrepancy.
Read the full abstract
Information Retrieval
From Language to Behavior: Scaling Sequence Transformers for Industrial Recommendation Ranking with Rec-Native Designs
Scaling Transformers has driven large gains in language modeling, but transplanting this to behavior-sequence modeling in production ranking is challenging: recommendation differs in signal quality, where behavior sequences are noisy, temporally irregular, and sparsely supervised, and in computation asymmetry, where each request scores many candidates against one shared user history under tight latency budgets. We propose ReST, a recommendation-native Transformer scaling framework. For signal quality, it introduces a sequence encoder with dual-gated attention, rotary positional and temporal embedding, stabilized residual normalization, and training-only auxiliary objectives. For computation asymmetry, it factorizes ranking into a heavy reusable encoder and a lightweight cross decoder with projection-free KV attention and token-specific parameterization, coupling user-level shared-prefix training with shared-prefix serving for compute-once, decode-many-times ranking. Across industrial and public benchmarks, ReST achieves higher accuracy and scales more consistently along sequence length, depth, and width, where LLM-style Transformer blocks saturate. A one-week online A/B test on a production advertising platform improves online AUC by 1.31% and lifts a core revenue metric by 11.93% within a 50 ms P99 budget; ReST has since been fully deployed in production, showing that behavior-sequence scaling remains a promising, under-exploited axis for production ranking.
Read the full abstract
Optics
Exceptional Points in Photonics: From Non-Hermitian Physics to Applications
Open photonic systems provide a versatile platform for non-Hermitian physics, enabling control over complex spectra, transport, and light-matter interactions. Exceptional points (EPs), at which eigenvalues and eigenvectors coalesce and the governing operator becomes defective, play a central role because they combine branch-point spectral topology, nonanalytic perturbative response, and controllable eigenstate conversion. This Review provides a unified framework for EP photonics by systematically distinguishing exceptional degeneracies according to the underlying operator, spectral variable, boundary conditions, and experimentally accessible observables. We discuss Hamiltonian EPs, absorbing EPs associated with scattering zeros, real-frequency scattering-matrix and Jones-matrix EPs, Bloch and Floquet EPs, and Liouvillian EPs in open quantum systems. We review their spectral topology, static and dynamical encircling, higher-order exceptional structures, and coexistence with bound states in the continuum, together with applications in sensing, lasing, coherent absorption, directional scattering, polarization and wavefront control, nonlinear optics, optical storage, nonreciprocal photonics, and quantum photonics. We also critically assess the current limitations, practical challenges, and future perspectives of EP-based photonic technologies, with particular attention to robustness, noise, scalability, and experimentally measurable performance.
Read the full abstract
Cosmology and Nongalactic Astrophysics
First tomographic measurements of the angular clustering and bias of photometric quasars from S-PLUS
We investigate the quasar clustering in the QuCatS photometric catalog, constructed from the fourth data release of the Southern Photometric Local Universe Survey (S-PLUS). The quasar sample for this study consists of $23,402$ quasars, spanning the redshift interval $1.1 \leq z_{\rm phot} \leq 2.6$, which we divided into four bins for our tomographic analyses. In each bin, we measured the two-point angular correlation function and compared it with theoretical predictions for the clustering of dark matter in the flat-$\Lambda$CDM cosmological model. We then estimate the effective linear bias in each bin that best fits the data, obtaining $b_{Q}(z_{\rm eff}=1.26)=2.17^{+0.30}_{-0.35}$, $b_{Q}(z_{\rm eff}=1.63)=2.67^{+0.36}_{-0.42}$, $b_{Q}(z_{\rm eff}=2.04)=3.29^{+0.52}_{-0.62}$, and $b_{Q}(z_{\rm eff}=2.38)=4.05^{+0.71}_{-0.86}$. These measurements are consistent with previous analyses in the literature. Moreover, the evolution of the bias agrees well with the $b(z)$ parameterization proposed by \cite{laurent2017}. Our measurements constitute the first cosmological study of the tomographic clustering of photometric quasars provided by the S-PLUS dataset, contributing to the mapping of the evolution of linear bias and paving the way for future analyses of data releases with larger datasets. In addition, our study also serves as a consistency test for the calibration procedure of the photometric redshift probability distributions of the quasars in the QuCatS catalog.
Read the full abstract
Analysis of PDEs
The $\alpha$-Limit Problem: Convergence of a Linear Degenerate Interface Transmission Problem
We study the singular limit of a family of linear degenerate interface transmission problems arising from a regularization procedure in the newly proposed Two-Parameter Diffuse Domain Method (DDM2p). For $\alpha>0$, the regularized problem admits a strictly convex variational formulation on $H^{1}(\Omega)$. In the limit $\alpha\to0$, the problem degenerates to a weakly coupled interface system with a nonstandard energy structure. To characterize the limit, we introduce a closed Hilbert subspace $\mathcal{H}\subset H^{1}(\Omega)$, defined through an auxiliary Helmholtz problem on an annular subdomain $\Omega_2\subset \Omega$, and identify the limiting energy functional $\mathcal{E}_{0}$ on $\mathcal{H}$. We prove that the regularized energies $\mathcal{E}_{\alpha}$ restricted to $\mathcal{H}$ $\Gamma$-converge to $\mathcal{E}_{0}$ in the strong $L^{2}(\Omega)$ topology, using the standard framework. Consequently, minimizers of $\mathcal{E}_{\alpha}$ converge to the unique minimizer of $\mathcal{E}_{0}$, which is shown to be equivalent to the solution of the limiting interface problem. We further prove a strong convergence $u_{\alpha}\to u_{0}$ in $H^{1}(\Omega)$ and establish a convergence rate of $O(\alpha)$. Numerical experiments in one spatial dimension exhibit first-order convergence, suggesting that the proved $O(\alpha)$ upper bound is sharp.
Read the full abstract
Software Engineering
Continuous Autonomous Refactoring: A Research Roadmap for AI-Driven Code Quality Maintenance
Large language models have shown promising capabilities in code refactoring, but existing approaches remain limited to method-level tasks. In this paper, we envision LLM-based refactoring as a continuous component of software maintenance rather than a tool invoked only for occasional manual refactoring. Under this vision, AI agents continuously monitor, evaluate, and improve codebases against explicit and evolving notions of software quality. We present a roadmap organized around five dimensions: the multi-objective optimization problem, quality definition and evaluation, multi-timescale integration of heterogeneous signals, architecture and design pattern, and trust in autonomous refactoring. We further identify integration into continuous delivery pipelines and cost considerations as cross-cutting concerns. For each dimension, we analyze the underlying challenges and pose open research questions. These dimensions define a research agenda for advancing autonomous refactoring from isolated code improvements to system-level quality maintenance.
Read the full abstract
Cryptography and Security
MutMem-V2: Cryptographically Authorized Mutation in Persistent Agent Memory Portable Verification and Reproducible Evidence
MutMem V1 introduced retention-preserving, cryptographically authorized mutation for persistent agent memory but did not provide a complete portable verification contract or clean-install reproduction path. MutMem V2 closes that publication gap without introducing a second memory engine. It specifies exact canonical bytes, domain-separated object and bundle commitments, mandatory recall-evidence membership and ordering, external trust anchors, identity epochs, revocation, authorization, request receipts, ordered disclosure, and three mutation terminal types. The released protocol contains 18 versioned object schemas, 39 recall vectors, 15 mutation vectors, and 37 closed recall failure reasons. Independent Node and Python implementations agree on verdict and primary reason for all 72 structural and cryptographic terminals; a production-conformance corpus agrees on 42/42 cases across 28 required classes. A clean Node v26.8.1 installation reaches first-boot, restart, and scheduler readiness with no experimental memories. A separately scoped 120-unit Canary experiment supports only explicit-marker traversal. Every public table regenerates from a self-hashed aggregate, and an independent verifier reconstructs the statistics and claim boundaries. Historical V1 empirical results remain historical. MutMem V2 supports claims about portable integrity, authorization, traceability, conformance, and reproducibility under stated assumptions; it does not establish semantic truth, universal robustness, or independent replication.
Read the full abstract
Group Theory
On commensurations of pro-$\mathcal{C}$ groups
We study commensurations of pro-$\mathcal{C}$ groups, where $\mathcal{C}$ is an extension-closed variety of finite groups. We characterize those commensurations of pro-$\mathcal{C}$ groups that can be extended to embeddings into a common overgroup, obtaining a pro-$\mathcal{C}$ analog of a recent theorem by Touikan. We also consider obtain a structural result on virtual retracts of pro-$\mathcal{C}$ groups.
Read the full abstract
Mesoscale and Nanoscale Physics
Interferometry reveals spin-singlet fractional quantum Hall edges in graphene
The edge modes of spin-singlet fractional quantum Hall (FQH) phases are manifestations of multicomponent topological order and SU(2) spin symmetry. An archetype is the spin-unpolarized state at $\nu$=2/3, whose edge is expected to host spatially coexisting yet counter-propagating charge and neutral spin modes. Despite efforts, its edge properties, including spin coherence and spin-charge separation, have remained elusive owing to the difficulty of resolving spins in FQH edge transport. Here, we develop a spin-sensitive probe of graphene FQH edges by using a p-n junction to interface a target FQH state with a probe integer quantum Hall (QH) state of opposite polarity. An Aharonov-Bohm (AB) interferometer forms along the interface, when the target and probe edge channels carry the same spin. We identify the spin-unpolarized and polarized edges of the $\nu$=2/3 states at lower and higher magnetic fields, respectively, through spin-dependent interference. A novel multiparticle AB interference between two electrons of opposite spin emerges from spin-charge separation and recombination on the unpolarized edge, featuring a spin swap. Our results establish edge transport as a probe of spin-singlet topological orders, with implications for parafermion platforms in graphene-superconductor hybrids.
Read the full abstract
Cryptography and Security
Position Matters: Feature Inversion Attacks in ViT Split Inference with Token Reduction and Shuffling
Vision Transformers (ViTs) are increasingly used in split-inference systems, where edge devices transmit intermediate token representations to a remote cloud. In this setting, token reduction lowers computation and communication costs, while token shuffling disrupts the spatial organization of the transmitted tokens, potentially limiting information leakage. However, their privacy benefits remain unclear against feature inversion attacks, which attempt to reconstruct the input from the transmitted embeddings. In this work, we show that, despite disrupting the spatial structure required by conventional reconstruction attacks, transmitted token embeddings retain substantial positional information. Based on this observation, we introduce the Spatially Aligned Reconstruction Attack (SARA), a unified pipeline that predicts token positions, restores their spatial layout, reconstructs missing embeddings using a feature-space masked autoencoder, and recovers the input image. Our results demonstrate that token shuffling provides only apparent privacy, as SARA largely reconstructs the original token organization. Token reduction offers stronger protection, but significant leakage persists when the retained tokens preserve sufficient semantic and positional information. Finally, we introduce a lightweight edge-side defense that removes positional embeddings and progressively adapts the edge-side transformer blocks through knowledge distillation. It substantially reduces attack performance against SARA, while preserving downstream task accuracy and requiring no changes to the cloud-side model.
Read the full abstract
Machine Learning
Multi-Head Self Attention is a Parameter Identification Mechanism
We prove that a multi-head scaled dot product attention can be viewed as a parameter identification strategy. The ratio of unidentified parameters to the total number of parameters scales like the reciprocal of the number of heads ($1/2 \to 1/(2H)$), meaning models with more heads are structurally more identified. A subtle side effect of the mathematics observation that attention can never be fully identified. Similarly we also show that some bias terms can have no effect on softmax-based attention layers in both the single- and multiple-head settings, though this is mostly a curiosity that should have a marginal effect on model size and model training/prediction efficiency. We also touch on modern improvements to transformers including RoPE and GQA from this perspective, illustrating how those as well can improve the ratio of “meaningful” parameters to all parameters. Simple numerical examples demonstrate that training can indeed involve updates that overlap model-invariant subspaces that arise from a lack of identification. As part of our experiments we use a “rebalancing” approach that can “fix” updates that overlap unindentified subspaces but do not try to present evidence this should actually be adopted. Instead we simply view our numerical results as exploring and confirming the theoretical results. As a whole we discuss a purely mathematical/statistical explanation, identification, for why specific architectural choices in transformers may have improved performance.
Read the full abstract
Logic
The Composition Lemma for $n$-dependence
We prove that a relation obtained by composing arbitrary functions of arity $\leq k$ with a relation definable in an $n$-dependent structure is $kn$-dependent. This confirms a conjecture of Chernikov and Hempel. We also demonstrate optimality of the result.
Nuclear Theory
Fundamental geometric limitations on disentangling nuclear-surface properties in relativistic heavy ion collisions
The extraction of the nuclear surface diffuseness from relativistic heavy ion collisions is limited by the intertwined responses of geometry-driven observables to surface diffuseness and intrinsic nuclear deformation. We investigate this limitation using event-by-event Monte Carlo Glauber simulations, focusing on the sensitivity of multiparticle correlations to the Woods–Saxon surface diffuseness $a_0$ in intrinsically deformed nuclei. We systematically examine the local correlations between $a_0$ and quadrupole and octupole deformation parameters, $\beta_2$ and $\beta_3$, and determine how these correlations affect the ability of different observables to constrain $a_0$. We find that observables dominated by elliptic geometry exhibit a strong response to quadrupole deformation, leading to a local $a_0$–$\beta_2$ degeneracy that substantially limits their sensitivity to the nuclear surface diffuseness. Triangular correlations provide a more independent response to the nuclear surface and therefore retain additional information on $a_0$, although their sensitivity can also be modified by intrinsic deformation. Extending the analysis to simultaneous quadrupole and octupole deformation shows that the local degeneracy and least-constrained directions depend on the nuclear configuration, demonstrating that the limitation on extracting $a_0$ is not described by a single global parameter correlation. We quantify these effects using multidimensional response maps, local sensitivity and information-geometric measures, and a Bayesian analysis of the resulting parameter constraints. The results clarify the fundamental limitations imposed by intrinsic multipole deformation on the determination of nuclear surface diffuseness from relativistic heavy ion collisions and identify multiparticle correlations that provide more independent information on $a_0$.
Read the full abstract
Quantitative Methods
On the interpretation of the kinetics of ligand-receptor binding
When the rates of ligand binding are measured by methods such as surface plasmon resonance, it is common practice to use the observed rate constants for the onset and offset of binding to estimate an equilibrium constant for ligand binding. If this agrees with the equilibrium constant found as the EC50 for binding at equilibrium, this is taken as validation of the measured rates. This is correct only when binding produces no conformation change in the receptor, and ligand binding follows a single exponential time course. Here, we investigate a simple 3 state model in which binding is followed by a conformation change in the receptor. Three special cases of this model in which the time course of onset and offset of ligand binding are close to being single exponentials are analysed. These cases are (1) when binding is much faster than the conformation change, (2) when the conformation change is much faster than binding, and (3) when the rates of ligand dissociation and receptor activation are both fast. It is concluded that the measured rates will often yield an estimate of the equilibrium constant for ligand binding that is close to the effective, or macroscopic, equilibrium constant, the EC50 found by measuring binding at equilibrium, which depends on both of the underlying microscopic equilibrium constants describing ligand binding and the conformation change. The exception to this conclusion is the case when binding is much faster than the subsequent conformation change, though the estimate of the equilibrium constant for ligand binding still depends on both of the underlying microscopic equilibrium constants.
Read the full abstract
Metric Geometry
A Projection Identity for Simplices Sharp Inequalities, Converse Results, and Affine Projections
We study a projection identity for a simplex in Euclidean space, written in terms of the frame operator of its unit edge directions. For a right simplex, the identity leads to a sharp family of distance inequalities and a complete description of equality. For a general simplex, the same formula is controlled by the spectrum of the Gram matrix through the Ky Fan principle. We prove converse results that characterise right simplices and determine the smallest number of projection subspaces needed to force orthogonality, together with an optimal quantitative estimate. We also treat affine projection subspaces and show how the original inequality for mutually perpendicular vectors fits into the same framework.
Read the full abstract
Signal Processing
Freshness-Aware Constrained Sensing-Aided Beam Prediction with Knowledge Distillation
Beam prediction leveraging environmental data reduces over-the-air beam training overhead. Existing frameworks, however, assume continuous access to fresh sensory data, an assumption that breaks down under sensing budget constraints or sensor failures. To make this more practical, this paper proposes a sensing-aided beam prediction framework that operates under an average sensing rate constraint. We incorporate the age of information (AoI) directly into the beam prediction pipeline as a synthetic input modality: the age of the most recently captured data is encoded and fused with visual features through a gating mechanism. This provides the predictor with explicit context information about input sensory data reliability. We formalize and examine three fixed sampling policies, accumulated, uniform, and randomized, under the average sensing budget. We further develop a knowledge distillation (KD) framework that operates as a robustness regularizer rather than a pure model compression method. In particular, the high-capacity teacher is trained unconstrained on fully sampled data, and its representational knowledge is transferred to a compact student deployed under the sensing budget. We conduct numerical experiments on the DeepSense 6G data set. The results show that AoI fusion nearly doubles top-1 accuracy at strict sensing budgets, and age-aware models achieve near-optimal top-3 accuracy with only 20% of the data. Furthermore, we find that the teacher training regime is a more consequential design choice than the distillation loss function.
Read the full abstract
Computer Vision
S$^2$Prune: Spatially Structured Visual Token Pruning for Multimodal Large Language Models
Visual token pruning reduces the inference overhead of multimodal large language models (MLLMs) by retaining only a subset of visual tokens. Existing methods usually select tokens based on importance or redundancy. However, we observe that these criteria produce stable spatial biases across inputs and do not always outperform simple Uniform Grid sampling, highlighting the value of broad spatial coverage. Motivated by this, we propose S$^2$Prune, a training-free pruning method that preserves spatial coverage while adapting token density to local image structure. We first divide the image into regions and assign at least one token to each region to preserve coverage. The remaining token budget is then distributed according to Laplacian variation, giving more tokens to regions with richer structure. We then use Early Representation Change (ERC), computed from the first decoder block, to select representative tokens within each region. We evaluate S$^2$Prune across diverse settings and two MLLM architectures. On Qwen2.5-VL-7B-Instruct, it achieves the highest average accuracy among the evaluated training-free pruning methods. With only 32 of the original 576 visual tokens, it still retains 79.3% of the full-model performance. Code is available at https://github.com/yuanyuanjia71-spec/S2Prune.
Read the full abstract
Materials Science
Geometry-Controlled Magnetic and Electronic Landscapes in Anisotropic van der Waals Materials
Electronic structure in van der Waals materials is commonly engineered through composition, strain, electrostatic gating and heterostructure assembly. Here we introduce geometronics, a concept in which substrate geometry locally reorients an anisotropic crystal, transforming homogeneous external perturbation into programmable magnetic and electronic landscapes. We demonstrate this concept using a bilayer of the antiferromagnetic semiconductor CrSBr transferred onto an inverted pyramidal nanoindentation, where the local crystal orientation with respect to the external magnetic field drives the coexistence of antiferromagnetic and ferromagnetic phases within a single continuous crystal. The resulting magnetic landscape creates a switchable excitonic potential well of up to 10–12 meV, directly visualised by spatially resolved spectroscopy. More generally, geometronics provides a universal route for deterministically programmed electronic and magnetic landscapes without modifying the material itself. It therefore establishes substrate topography as a new design principle that exploits the intrinsic anisotropy of layered van der Waals materials.
Read the full abstract
Cryptography and Security
What's in Your Agent's Context? Context Privilege Escalation Attacks against AI Agent Harness
Real-world, high-profile AI agent harnesses often rely on vendor-proprietary or opaque designs for context assembly, leaving the sources and underlying logic of assembled context poorly understood and the resulting security risks largely unexplored. In this paper, we present the first systematic analysis of context assembly designs in real-world AI agent harnesses. We study and uncover how an agent harness is designed to collect and assemble context from diverse sources, and identify a set of practical attack vectors arising from these designs. Our analysis brings to light two novel categories of attacks in the context assembly of real-world harnesses: (1) MessageRole Context Privilege Escalation (M-CPE), which occurs when attacker-controlled content originating from a low-privileged context is incorporated into a higher-privileged message role. (2) Cross-Scope Context Privilege Escalation (X-CPE), which occurs when attacker-controlled content persists beyond the context in which it was introduced. We performed a systemic security analysis of the CPE attacks against 12 real-world agent harnesses, including Claude Code and Codex. The resulting consequences include full agent compromise, remote code execution, denial of service, and manipulated tool or skill invocations, etc.