Computation and Language
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Wednesday 9 September
We introduce BuzzASR, a collection of language-specialized fine-tuned Whisper models adapted for automatic speech recognition (ASR) in 102 languages. Large end-to-end Transformer-based ASR models such as Whisper have revolutionized ASR, but most prominent models are highly multilingual. As a result, these models often perform poorly on languages less well-represented in their training set. While it has long been known that effective language adaptation can be achieved through simple fine-tuning on monolingual data, this strategy has only been applied to a small number of languages. We massively scale up this simple approach to 102 languages covered in the FLEURS dataset, while also implementing a more complex language adaptation strategy that integrates monolingual tokenizer replacement and data augmentation using text-only fine-tuning. BuzzASR models outperform Whisper-large-v3 on 77 out of 102 languages, reducing character error rates (CER) by a factor of over 2.8 on average. Our models achieve state-of-the-art CER among open-source systems on 27 of 102 languages on the combined FLEURS and Common Voice test set. Our tokenizer replacement strategy yields an average 3.3x improvement in compression rate (characters per token) over Whisper's multilingual BPE, with gains of up to 21.7x. We release all models, code, and detailed results: https://lemn-lab.github.io/buzz-asr
Cryptography and Security
Arbitrary Cipher Attacks Against Large Language Models Do Not Require Fine-Tuning
Large language model safety and security research is preoccupied with, among other things, detecting and preventing jailbreak attacks: alignment bypasses that allow an adversarial user to elicit unwanted or harmful outputs from models. Arbitrary cipher, or covert communication, attacks are one such type of jailbreak and have previously been demonstrated against the fine-tuning APIs of commercial models. In these attacks, target models are trained on a corpus of encrypted harmful questions and responses and subsequently respond to harmful requests through the learned encryption scheme. In this paper, we show that newer frontier models do not require fine-tuning to acquire cipher-based communication skills. Instead, they can learn these skills through prompting and, when necessary, through in-context learning. Furthermore, model alignment is significantly weakened or entirely bypassed when communication occurs through the learned cipher. To the best of our knowledge, this constitutes a novel attack vector against commercial black-box large language models. We demonstrate successful jailbreaks against frontier models developed by Anthropic, Google, and OpenAI. Our attack bypasses commercial harmfulness classifiers because harmful content is encrypted and therefore appears as nonsensical text or gibberish.
Read the full abstract
Computation and Language
TEFM: Token-Efficient Faithful Modeling for Structured Data
In this paper, we solve two fundamental obstacles in applying LLMs to critical domains: token efficiency and faithfulness. To address both constraints jointly, we present TEFM (Token-Efficient Faithful Modeling), a framework designed for structured data analysis in critical domains. TEFM achieves token efficiency by compressing lengthy structured observations into compact Behavioral Code tokens, dramatically reducing token consumption with minimal information loss. Moreover, TEFM enables faithful rationalization through a dual-fidelity objective that jointly optimizes code-level reconstruction and prediction-level fidelity, identifying minimal sufficient feature subsets grounded in input data. Comprehensive experiments across various domain datasets and model backbones (Qwen3, Gemma-2, Phi-4) show that TEFM achieves competitive classification accuracy with dramatic token reduction (approximately 1% token retention in clinical and 2% in security domains) while producing faithful rationales.
Read the full abstract
Cryptography and Security
An Efficient and Effective Agentic Group Shilling Attack on Recommender Systems
Recommender systems have become core infrastructure for modern online platforms, personalizing content at scale and strongly influencing what users see, click on, and purchase. However, this dependence on user interaction also exposes them to shilling attacks, where malicious actors can inject fake profiles to distort item rankings and control visibility. Existing attacks often rely on target-specific fine-tuning or fixed profile templates, making them either difficult to adapt to different victims or easier to detect. To overcome these limitations, we propose the Agentic Group Attack System (AGAS), a coordinated shilling framework where a central Coordinator directs a group of role-switching worker agents to adaptively promote a target item across different victim families. The Coordinator dynamically adjusts the strategy when progress stalls or suppression signals increase, while workers pursue a shared objective and switch between active and inactive roles to avoid repetitive patterns. Under the same attack budgets and evaluation protocols, AGAS consistently surpasses strong baselines in target promotion while better preserving benign recommendation quality, weakening representative detectors, and achieving higher efficiency than prior attacks. These findings also emphasize that defending recommender systems may require mechanisms that can handle adaptive shilling campaigns, not just isolated fake-profile injections. Our code is available at https://github.com/phkhanhtrinh23/AGAS.
Read the full abstract
High Energy Physics (Theory)
On general background of quantum non-invertible symmetry in 2D
We study the dual quantum symmetry $\mathrm{Rep}(G)$ in the two-dimensional theory $\widetilde{\mathfrak{T}}_{\mathrm{Rep}(G)}$ obtained by gauging a non-abelian symmetry $G$ of $\mathfrak{T}_G$, where for Lie groups $G$ the gauging is understood as flat gauging. We develop a general framework for computing partition functions of $\widetilde{\mathfrak{T}}_{\mathrm{Rep}(G)}$ in arbitrary non-invertible symmetry backgrounds, represented by topological defect networks of $\mathrm{Rep}(G)$, in terms of the partition functions of the original theory $\mathfrak{T}_G$. We also derive the inverse transformation, expressing partition functions of $\mathfrak{T}_G$ in $G$ backgrounds in terms of those of $\widetilde{\mathfrak{T}}_{\mathrm{Rep}(G)}$. We test the framework in several examples, including finite groups with multiplicity-free and higher-multiplicity fusion rules, and discuss a formal extension to compact Lie groups, focusing on $\mathrm{Rep}(SU(2))$.
Read the full abstract
Computational Geometry
The Hyperbolic Surface Distance, Diameter, and Dirichlet Problems
Despite the prominence of hyperbolic surfaces in mathematics, basic algorithmic questions about them, even computing the distance between two points, have remained open, leaving many features of these surfaces inaccessible. The classical machinery assumes a polyhedral structure absent on a smooth surface. We remove these obstacles. We begin with an efficient $O(g^2)$ algorithm for the distance between two points, where $g$ is the genus of the surface. Building on it, we obtain an $O(g^2 \log g)$ method for answering distance queries from a fixed source and, as a consequence, for recentering a Dirichlet domain around an arbitrary point. This understanding of distances on the surface then lets us approximate the diameter to within any $\eps$ in time $O(g^3 \log g / \eps^2)$. We further show that the diameter, a single real number encoding a great deal about the surface, is exactly computable. Its hyperbolic cosine is an algebraic number over the field encoding the coefficients of the hyperbolic isometries defining the surface.
Read the full abstract
Mathematical Physics
Bargmann-Fock Representation and Global Estimates for the Linearized Hard-Sphere Boltzmann Operator
We study the linearized Boltzmann collision operator for three-dimensional hard spheres and reorganize the complete operator so that its angular and radial structures are explicit before any Sonine truncation. From the Carleman representation, the collision integral becomes a superposition of orthogonal translations; in Hermite/Fock and Bargmann form these generate a single analytic coherent-state kernel for all Hermite collision brackets. Rotational covariance separates fixed angular-momentum sectors, each reducing exactly to a one-variable radial Bargmann space with an $\mathfrak{su}(1,1)$ structure. The radial operator is $\mathfrak A_\ell=\kappa\pi G(\mathcal J_\ell)-\mathfrak C_\ell$, where $\mathcal J_\ell$ is an explicit Laguerre–Jacobi operator and $\mathfrak C_\ell$ is compact. A unitary intertwiner identifies this realization with the classical Burnett/Sonine coordinate, recovering the Burnett matrix and standard stress and heat-flux corrections. For the untruncated operator, the compact gain gives only compact corrections, while the loss sets the common essential spectrum $[\nu_{\min},\infty)$. Radially squeezed $\mathfrak{su}(1,1)$ coherent packets yield threshold Weyl sequences in every fixed angular sector. With $\ell\to\infty$ and squeezing scaled as $\ell^{-1}$, they furnish full-space Weyl sequences throughout the essential band interior, with $\langle j\rangle\sim\ell^2$, revealing quadratic continuum corridors in the $(\ell,j)$ plane. At fixed $\ell$ and $j\to\infty$, the matrix tail approaches a universal Toeplitz operator whose zero at $\theta=\pi$ is the local signature of the same threshold escape. Thus the Fock reformulation provides exact radialization and constructive control of the global spectral geometry.
Read the full abstract
Machine Learning
A Statistical Approach to Estimating Sample Size of Machine Learning Models
Sample size determination for machine learning (ML) prediction models is challenging because conventional power analysis typically requires the predictor-outcome relationship and effect structure to be specified a priori. Nonlinear ML models learn complex prediction surfaces that do not admit straightforward analytical power calculations. We propose a framework that approximates nonlinear ML models with localized linear representations and estimates sample size requirements by evaluating statistical power across these local regions.
Read the full abstract
Numerical Analysis
A Convergent Reaction-Drift-Diffusion Master Equation with Interaction Potentials
We extend the convergent reaction-diffusion master equation (CRDME) framework on unstructured grids to include two-body interaction potentials between particles. Such interactions are essential for modeling systems in which particles experience electrostatic repulsion, volume exclusion, or van der Waals attraction. Our approach discretizes the weak form of the forward equation for a particle-based stochastic reaction- drift-diffusion model to obtain a new CRDME by combining two complementary techniques: an edge- averaged finite element method with mass-lumping quadrature for spatial transport terms that include potential interactions, and a mix of finite volume and mass-lumping quadrature approximations for reaction terms. This yields transition rates for a spatial jump process of particles hopping between dual mesh voxels and undergoing reactions. The resulting spatially discrete CRDME retains key physical properties of the continuum system in closed domains. It preserves the Gibbs-Boltzmann equilibrium distribution in the absence of reactions, and satisfies detailed balance of both drift-diffusion and reactive fluxes for systems with reversible reactions. Numerical comparisons with Brownian dynamics simulations validate our method, and we demonstrate empirical second-order convergence as the mesh is refined.
Read the full abstract
Quantum Physics
Sub-extensive non-stabilizerness in the Dyck-Fredkin spin chain
The stabilizer Rényi entropy is a quantitative measure of non-stabilizerness, or magic, and has typically been found to scale extensively with system size $N$ (i.e., $\Theta(N)$) for a variety of many-body quantum states. In this note, we study the stabilizer Rényi entropy of the ground state of the spin-$\frac{1}{2}$ Dyck-Fredkin chain and its $t$-deformation, a local frustration-free model with unusual spectral-gap scaling. Exploiting the combinatorial structure, we carry out numerically exact finite-size calculations, which indicate asymptotic behavior depending on $t$: $\Theta(N)$ for $t<1$, $\Theta(\log N)$ at $t=1$, and $\Theta(1)$ for $t>1$. The scaling at $t=1$ could be another manifestation of the unconventional criticality of the model, while the contrast with the behavior of the entanglement entropy suggests that non-stabilizerness might provide a new window into quantum many-body systems.
Read the full abstract
Tuesday 8 September
Quantitative Methods
Physics-informed and event-aware forecasting of ammonia emissions in aerobic chicken manure composting bioreactors: A PINN-PatchTST-GPR framework with symbolic dynamic extraction
Predicting ammonia (NH3) concentrations during aerobic chicken manure composting provides critical feedforward signals for intelligent aeration. However, the concurrence of long-term biochemical evolution and intermittent forced-aeration pulses poses significant modeling hurdles across steady baselines, transient surges, and risk quantification. Existing data-driven models lack physical consistency, while mechanistic kinetic models struggle with environmental disturbances. To bridge this gap, we propose PINN-PatchTST-GPR, a physics-informed neuro-symbolic forecasting framework. Validated on a 28-day composting campaign across six vertical bioreactors, the framework couples a channel-independent PatchTST backbone with macroscopic kinetic constraints, dynamic physical bounds, and a Gaussian process regression (GPR) event specialist. Multi-channel sensor streams drive discrete differential equation residuals and non-negativity bounds, strictly ensuring physical plausibility. A block-split conformal prediction module yields nominal 90% distribution-free prediction intervals, while PySINDy and PySR extract low-complexity symbolic equations for quiescent and surge regimes. The model achieves an overall RMSE of 0.8256 +/- 0.0132 ppm, R2 of 0.9817 +/- 0.0006, Epeak of 1.6793 +/- 0.0725 ppm, and Etime of 0.7 +/- 0.2 min. For uncertainty quantification, it attains PICP90 of 99.792 +/- 0.047%, MPIW of 6.951 +/- 0.128 ppm, and IS0.1 of 6.968 +/- 0.128 ppm. Symbolically, PySINDy yields an RMSE of 0.4624 +/- 0.0758 ppm (R2 = 0.9896), while PySR yields 2.4084 +/- 0.0864 ppm (R2 = 0.9523). Biologically, the 20-min aeration strategy minimizes cumulative NH3 emissions. Overall, PINN-PatchTST-GPR harmonizes global forecasting, localized transient responsiveness, and risk quantification for precision composting.
Read the full abstract
Computation and Language
Bridging Semantics and Utilities: A Controllable and Scalable Market Simulation
Large language models(LLMs) have recently been used to construct generative agents for financial market simulation. Existing approaches typically prompt each agent separately with public market information and a private investor profile, and ask the model to directly generate trading actions. In this agent-wise design, the model repeatedly interprets information shared across the population, while individual behavior is controlled through textual descriptions of investor preferences and traits. This creates two structural difficulties: continuous traits such as risk sensitivity do not reliably induce ordered behavioral changes, and repeatedly processing the same public information causes inference costs to grow with the number of agents. We propose SUM, a Semantic-Utility Model that separates cohort-level semantic perception from agent-level utility-based decision-making. A language model interprets public market information for each investor type and produces semantic signals shared within the corresponding cohort. Individual agents then combine these signals with their private states, preferences, and constraints to generate orders through utility-based decisions. Experiments in an A-share limit-order-book market show that SUM provides more stable and monotonic behavioral control than prompt-based agents while substantially reducing repeated language-model inference as the population grows. Closed-loop simulations further reproduce directional market responses, volatility persistence, differentiated investment styles, and persistent wealth heterogeneity. These results establish SUM as a controllable, scalable, and semantically informed framework for financial market simulation.
Read the full abstract
Programming Languages
pfoqcompiler: Efficient Compilation of Quantum Control in Recursive First-Order Quantum Programs
Recursive quantum control allows for simple and concise descriptions of algorithms. However, a naive circuit compilation sequentializes the different branches of the quantum if-statement and results in an exponential increase in circuit depth. Existing techniques can parallelize these branches, but do so at the cost of an exponential number of ancillary qubits. We present pfoqcompiler, a compiler targeting Qiskit for the first-order quantum programming language FOQ that exploits program structure to merge compatible recursive procedure calls across quantum branches. We introduce weaker syntactically-checkable conditions for well-foundedness, recursive halving, and uniformity, while preserving previously demonstrated complexity characterizations. For weakly well-founded bounded-width programs, the compiler produces polynomial-size circuits; weakly recursively halving programs additionally yield polylogarithmic depth, while weak uniformity removes the asymptotic overhead of branch sequentialization.
Read the full abstract
Statistics Applications
Manipulation-Resistant Oracles via Continuous Peer Prediction
Blockchain oracle networks, including decentralised physical infrastructure networks (DePINs), report physical measurements from a source that can manipulate the underlying process. A carbon project operator, for example, can dry a site before measurement and be paid for soil organic carbon (SOC) that was never stored. Peer prediction covers discrete signals under a strategic source, or continuous reports under an honest source, but not both at once. We close that gap for Gaussian reports. Strict propriety of the continuous ranked probability score (CRPS) settles the reporting stage alone. Our result concerns the source stage. A report-only settlement rule whose honest-source payoff is non-decreasing in the true value identifies and deters source manipulation if and only if the sensors respond to that manipulation by different amounts. When they respond identically, manipulation is observationally equivalent to a genuine rise in SOC, and CRPS translation equivariance leaves every peer score unchanged, so no such rule deters the source. Necessity follows from invariance rather than from the signal enumeration available in the discrete case. The source-stage estimator is ordinary weighted least squares, and the contribution is the characterisation around it. With eight calibrated SOC models on 37,280 real European soil spectra, shift-naive aggregators over-credit by up to 57% under a calibrated moisture perturbation while the joint estimate stays near-unbiased. The guarantee holds on the equilibrium path for non-colluding reporters and inherits the models' own calibration error. The project page is at https://rqcker.github.io/crps-wls.
Read the full abstract
Computational Physics
Representation-split Carleman lattice Boltzmann with near-unit step success probability on quantum computers
The Carleman lattice Boltzmann method replaces the quadratic collision of the lattice Boltzmann equation by a linear map on a lifted state. Existing block encodings of this map succeed with probability $10^{-5}$ to $10^{-2}$ per step, so that the probability of a multi-step evolution decays exponentially with the number of steps. This study develops multi-relaxation-time Carleman lattice Boltzmann, a representation split in which the Carleman-lifted collision is performed in a moment basis on the amplitude-encoded state, while streaming remains a permutation in population space. The level-2 hierarchy closes under collision but not under streaming, which generates the work of pressure, advection and stress on the momentum flux. The state therefore remains the full pair array, and the moment representation is an orthogonal change of basis of the population one, with identical dynamics and truncation error. The operator to be block encoded changes with the basis, and with it the success probability of a quantum step. In the weighted Hermite basis with the amplitude encoding $f_i/\sqrt{w_i}$, the relaxation is a diagonal operator of norm one, and the level-2 collision stage becomes this diagonal plus a sparse coupling from products of conserved moments. A structured block encoding of this form raises the single-step success probability on the D2Q9 lattice from $10^{-2}$ for the best population encoding to 0.98. With the pair sector in relative coordinates and scaled, one step requires $O(\log^2 N)$ Toffoli gates and is verified by statevector simulation. The encoding and readout overhead of a $T$-step run is then polynomial in $T$ and $N$ rather than accumulating a factor $p^T$. For resolved flows, the remaining exponential loss in coherent multi-step Carleman lattice Boltzmann evolution is set by the non-equilibrium norm relaxed by dissipation rather than by the block encoding.
Read the full abstract
Metric Geometry
A 0.41275 lower bound for the three-dimensional Blaschke-Lebesgue problem
We prove that every convex body $K\subset\mathbb{R}^3$ of constant width $w>0$ satisfies \[ \operatorname{Vol}(K)\ge \frac{1651}{4000}w^3=0.41275\,w^3. \] This improves the certified lower bound $0.411040473721188\ldots\,w^3$ announced by Hyra and the earlier spectral bound $4\pi w^3/33$ of Nishioka. The main new ingredient is a contact-tetrahedron estimate near equality in Jung's circumradius inequality. Four balanced contacts of a minimum circumscribed ball form a near-regular tetrahedron; constant width then forces, beyond the tetrahedron itself, spherical regions over its faces, spindle sectors along its edges, and additional face-edge sectors. These regions can be selected with pairwise disjoint interiors and quantified by the total squared-edge deficit. A sharp convexity argument relates that deficit to the circumradius, and a symmetrization argument controls the four face contributions with only quadratic loss. This gives the analytic bound $\operatorname{Vol}(K)>0.41275318$ near the Jung endpoint. For the complementary circumradius range we use Hyra's radial transport/tangent-majorant mechanism together with closed-form contact and antipodal profiles. The antipodal contribution has an elementary primitive. For the contact contribution, a second-order concavity estimate converts the angular Jacobian into an explicit rational lower density; that density is monotone in the circumradius. The whole intermediate range therefore reduces to a single endpoint calculation, giving $\operatorname{Vol}(K)>0.41275168$. For smaller circumradius the tangent-majorant inequality with its nonnegative gap discarded already gives $\operatorname{Vol}(K)>0.41308109$. The proof is entirely analytic and requires no computer-assisted certificate.
Read the full abstract
Rings and Algebras
Quasi-derivations of Block Lie algebras $\mathcal{B}(q)$
We compute quasi-derivations of Block Lie algebras $\mathcal{B}(q)$. In particular, we prove that $\mathcal{B}(q)$ has various nontrivial quasi-derivations, i.e., quasi-derivations different from derivations and $\delta$-derivations. As corollaries, we obtain a complete description of $\delta$-derivations, $\delta$-biderivations, and transposed $\delta$-Poisson structures on $\mathcal{B}(q)$ for $\delta\notin\{0;1\}$. In particular, for $\delta\notin\{0;1\}$, $\mathcal{B}(q)$ up to isomorphism admits only one nontrivial transposed $\delta$-Poisson structure if $q\in\mathbb Z$, and $\mathcal{B}(q)$ does not admit nontrivial transposed $\delta$-Poisson structures if $q\notin\mathbb Z$.
Read the full abstract
Statistics Theory
Entropy-Based Goodness-of-Fit Tests for Generalized $\kappa$ -Gaussian Distributions
We introduce this class of multivariate Kappa-Gaussian distributions by maximizing a type of Kaniadakis $\kappa$ entropy subject to certain first- and second-moment constraints. This differs slightly from the approach of Rényi and Tsallis, because the $\kappa$-entropy arises from a relativity-consistent deformation of the exponential and logarithmic functions and yields a family of distributions whose tails follow a power law, a family that actually reverts to the standard Gaussian distribution as $\kappa \to 0$. We then create a $k$-nearest-neighbor (NN) estimator for $\kappa$-entropy and describe it using the difference between two NN functionals involving sums of powers. Next, we establish $L^2$-consistency and asymptotic normality, using the well-known subadditive Euclidean functional and moment expansion framework for NN entropy estimators. Furthermore, using a scaling identity for the $\kappa$-exponential, we derive a closed-form formula for the entropy of the $\kappa$-Gaussian distribution itself. From this, we determine the exact threshold value $\kappa < 2/(m+2)$ at which the distribution preserves finite variance; this is analogous to the “degrees of freedom $>2$” condition in the Student's-$t$ distribution. With these components, we construct a non-parametric goodness-of-fit test and then verify it through simulations. Finally, we mention the Sharma–Mittal entropy, which combines the Rényi and Tsallis entropies as a natural direction for future work.
Read the full abstract
Combinatorics
The additive-square-free spectrum below 139
For a finite real alphabet $T$, let $g(T)$ be the supremum of the lengths of words over $T$ containing no two adjacent nonempty blocks of equal length and equal sum. We classify all four-element alphabets with $g(T)<139$. Up to affine equivalence, these are the known endpoint-sum family and seven isolated classes. Exhaustive finite searches give the isolated values, while 77 explicit words prove that every remaining real alphabet admits a word of length 139. The latter argument uses exact intersections of finite unions of lines, so it covers irrational parameters as well. The cutoff is sharp because $g(\{0,2,3,7\})=139$. We also obtain a sharp cutoff of 167 within the family $\{0,1,x,x+2\}$, $x>1$. The required supplement prints the word certificates and specifies the complete-search evidence.
Read the full abstract
Risk Management
Modelling Robust Lending Decisions under Climate Scenario Ambiguity: A Minimax-Regret Framework with NGFS Short-Term Scenarios
This paper develops a public-data framework for evaluating incremental bank lending when plausible climate scenarios imply different sector credit outcomes but defensible scenario probabilities are unavailable. The calibration combines 2025 Shared National Credit industry commitments, January 2026 U.S. industry leverage, interest coverage and equity volatility from Damodaran, and official 2027 sector-level probability-of-default adjustments from the NGFS CLIMACRED model. One-year Merton probabilities provide independent baseline credit-risk proxies for 16 non-financial sectors representing \$5.2808 trillion of commitments. Minimax regret is compared with symmetric scenario weighting and maximin under two principal decision scores: expected loss and a credit-compensation payoff using Damodaran synthetic spreads. DAPS raises the commitment-weighted one-year PD from 0.0592% to 0.0897%. At a 12.5% sector cap, minimax regret reduces maximum expected-loss regret by 29.1% relative to symmetric scenario weighting; under the credit-compensation score the reduction is 69.1%. Sensitivity to concentration limits, crosswalk weights, Merton inputs and the climate-PD transport shows that the value of minimax regret is conditional on scenario-specific sector rankings differing near binding portfolio constraints. The contribution is an operational decision-fragility framework for financial institutions and supervisors, not evidence about realized bank behavior or universally optimal climate lending.
Read the full abstract
Probability
Exact Universality of Online Discrepancy
We study online vector balancing with $N$ random vectors in $\mathbb{R}^M$ revealed sequentially, where each vector must be assigned an irrevocable sign upon arrival. The goal is to minimize the expected $\ell^\infty$ norm of the final signed sum. For i.i.d. entries with mean zero, variance one, and a finite fourth moment, we prove that, as $M/N\to\alpha\in(0,\infty)$, the optimal value divided by $\sqrt N$ converges to a limit $R_\alpha$ independent of the entry distribution. This limit is the stochastic control value identified for Gaussian inputs by Fiedler, Jackson, Lacker, and Niles-Weed. In particular, it determines the exact asymptotic optimum for Rademacher inputs. For every $\kappa>R_\alpha$, we construct a randomized online algorithm whose final signed sum has $\ell^\infty$ norm at most $\kappa\sqrt N$ with high probability; for $\kappa<R_\alpha$, every online algorithm has vanishing success probability. Consequently, the online threshold of the symmetric binary perceptron is universal at every positive margin. The main step is a coupling that transfers Brownian controls to non-Gaussian inputs, while truncation controls rare large entries.
Read the full abstract
Machine Learning
Uncertainty-Aware Learning from Multi-Expert Interval Targets
Many machine learning (ML) applications rely on expert labels, and qualified experts may provide different but plausible interpretations of the same observation. Such variation across expert labels may reflect genuine disagreement or ambiguity rather than annotation error. When individual experts additionally report intervals rather than exact values, the supervision contains two distinct sources of label uncertainty: within-label imprecision and between-expert variation. Existing methods treat these forms separately: multi-expert approaches collapse labels to a consensus, interval-target methods often yield a single prediction, and predictive-uncertainty methods rarely validate their uncertainty estimates against observed expert disagreement. To address this problem, we propose an approach that preserves individual expert intervals, separates within-label imprecision from between-expert variation, and validates the corresponding predictive uncertainty components. First, heterogeneous label vocabularies are harmonized into a common probabilistic label space, separating encoding differences from expert judgement. Second, individual label intervals are retained and modeled with a mixture of Beta distributions trained using a proper Cramér-distance objective, preserving distinct expert-reported labels. Third, we decompose predictive uncertainty into within-component, between-component, and model uncertainty, and evaluate whether these components correspond to within-label uncertainty, between-label uncertainty, and model error, respectively. Because this correspondence is not guaranteed, we introduce decomposition matching, which aligns the predictive components to their intended label-side sources. On sea-ice concentration the model reduces MAE by 31% over hard labels and outperforms aggregation, interval-distribution, and interval-regression baselines.
Read the full abstract
Machine Learning (Statistics)
Weighted Data Selection: Sharp Upper-Half and Five-Dimensional Laws
How much risk does a small reweighted training support retain? For finite weighted least squares with the minimum-norm learner, we prove the exact law $\Gamma_d(n)=3-n/d$ throughout $\lceil3d/2\rceil\leq n\leq2d-1$. The guarantee covers every observed feature rank and uses selections that preserve the full feature span. Balanced simplex anchors reduce dimension; positive-weight lifting and independent-line compression close the risk bound. Shifted coordinate pairs attain the matching lower bound. The complete dataset-level upper bound and sharpness construction are verified in Lean 4. At the smaller budget $(d,n)=(5,6)$, we also prove $\Gamma_5(6)=11/5$, matching the simplex-block prediction from $5=3+2$ over arbitrary interacting configurations. Circuit covers, comparison second moments, and circuit-plane probabilities give the sharp excess $6/5$, while polar-face geometry resolves shared rank-three circuits. The general simplex-block frontier connects these laws within the intermediate-budget selection problem.
Read the full abstract
Materials Science
Transferable Graph Neural Network Surrogates for Molecular Dynamics Across Crystal Symmetries
We present a transferable graph neural network (GNN) surrogate framework for molecular dynamics (MD) that directly predicts atomic displacements and propagates atomistic configurations without explicit force evaluation or numerical time integration. The central objective of this work is to establish whether a common GNN formulation can represent atomic dynamics across materials with fundamentally different crystal symmetries and coordination environments. The same network architecture, feature representation, graph construction, and training protocol are applied without symmetry-specific modification to face-centered cubic (FCC) aluminum, body-centered cubic (BCC) iron, and hexagonal close-packed (HCP) magnesium. Across these distinct elemental and crystallographic systems, the framework achieves position-prediction errors on the order of 10^-4 Angstrom^2 and supports stable autoregressive propagation to nanosecond time scales. The predicted trajectories preserve thermodynamic stability, characteristic coordination-shell structure in the radial distribution functions, and temperature-dependent mean-squared-displacement behavior. In particular, the framework captures the closely spaced coordination shells of BCC iron and the anisotropic coordination environment of HCP magnesium without introducing lattice-specific representations. These results demonstrate that direct GNN-based atomic propagation can be formulated as a transferable framework across different elements, lattice symmetries, and coordination geometries, providing a pathway toward generalizable surrogate models for accelerated molecular dynamics.
Read the full abstract
Combinatorics
Periodic Joint Co-tiles of Independent Tiles
For every $d\geq2$, we prove that $d-1$ independent finite tiles in $\Z^d$ admit a fully periodic joint co-tile whenever they admit a joint co-tile. Here independence means that the tiles contain the origin and every choice of one nonzero vector from each tile is linearly independent. This removes property $(\star)$ from the periodic existence theorem of Meyerovitch, Sanadhya and Solomon and recovers Bhattacharya's theorem when $d=2$. More generally, we prove periodic existence for any finite family admitting a joint co-tile whose indicator is a fully periodic real function plus finitely many bounded real functions, each invariant under a subgroup of rank $d-1$. Extending the Boolean normal-form method of Greenfeld and Tao, we obtain a configuration in the translation orbit closure whose restrictions to cosets of a finite-index lattice are either invariant under a rank-$(d-1)$ subgroup or given by a three-term affine fractional-part formula with density one half. The tiling equations pair the restrictions given by this formula and determine a finite bipartite graph. A two-coloring replaces those restrictions by constants while preserving every tiling equation. We deduce a characterization of periodic tilability by independent companions and a decision algorithm for families containing an independent $(d-1)$-subfamily, with the dimension as part of the input.
Read the full abstract
Cryptography and Security
Characterizing and Codifying Malware Sophistication
'Sophisticated' is widely used to describe malware, yet it lacks a consistent definition within academic literature. While existing software quality and complexity metrics offer some insight into malware structure, they do not capture the broader adversarial and operational traits that contribute to real-world threat potential. This paper presents a systematization of existing approaches for assessing malware quality using static binary analysis. We define malware sophistication through a quality-focused lens by reinterpreting select characteristics from the ISO/IEC 25010 software quality standard, including reliability, maintainability, flexibility, and security, and evaluating their applicability to malware binaries. We identify which characteristics are both relevant to malware and measurable through static analysis, forming the basis for future frameworks that consistently assess malware sophistication when source code is unavailable or dynamic execution is infeasible.
Read the full abstract
Computer Vision
DramaAgent: Agentic Storytelling Video Generation
Recent diffusion and autoregressive models have substantially improved text-to-video generation, yet producing coherent long-form story videos with consistent characters and aligned audio remains challenging. Existing methods often suffer from narrative drift, unstable character identity, weak cross-scene continuity, and audio-visual mismatch over extended sequences. We propose DramaAgent, a hierarchical, agentic, and model-agnostic framework for long-form text-to-video-and-audio generation. Rather than improving the underlying video backbone itself, DramaAgent introduces an upper-level control layer that decomposes generation into story planning, persistent character conditioning, scene-wise synthesis, and reflection-guided targeted repair. The framework maintains reusable story and character states across scenes, diagnoses failures such as identity drift, missing scene semantics, temporal discontinuity, and cross-modal mismatch, and repairs problematic clips in a stage-specific manner. Experiments across multiple video generation backbones show that DramaAgent improves long-horizon coherence, character consistency, narrative fidelity, and scene-level audio-visual consistency over direct generation and strong baselines. These results suggest that hierarchical agentic control is a practical direction for controllable long-form audiovisual generation. Code: https://github.com/AIGeeksGroup/DramaAgent. Website: https://aigeeksgroup.github.io/DramaAgent.
Read the full abstract
Computation and Language
FACET at WMT 2026 Automated Translation Quality Evaluation Task
Different error types in machine translation require different evidence. Whether meaning is preserved can be judged only against the source, while whether the target is well-formed, or whether it names one entity consistently, can be judged from the target alone. We present FACET, our reference-free submission to the WMT26 Automated Translation Quality Evaluation Task, which decomposes evaluation into Fluency, Accuracy, and Consistency passes and gives each pass only the context its error type requires. A single fixed model is prompted three times, and the merged error spans yield the three task outputs, error spans, quality scores, and error-free labels, with no trained components. We also submit FACET-C, which omits the Consistency pass. Without gold labels, we characterize the predictions of FACET. Its system rankings place post-edited human translation first, and the Consistency pass changes about a tenth of segment scores while leaving the ranking nearly unchanged.
Read the full abstract
Machine Learning
One Mastery Threshold Does Not Fit All Knowledge Tracing Models
Tutoring systems use mastery thresholds to decide when students can stop practicing and advance, but the same numerical threshold can lead to very different decisions when the underlying knowledge tracing (KT) model changes. We examine six KT models across four public educational datasets and evaluate 12 thresholds from 0.50 to 0.99 using post-advancement performance, advancement coverage, practice burden, and disparities across prior-performance groups. We also identify thresholds that balance performance, extra practice, and advancement under 30 predefined instructional settings. Bayesian Knowledge Tracing (BKT) is relatively insensitive to threshold changes, while neural models become much more selective as thresholds increase. This partly reflects different model outputs: BKT estimates latent mastery probability, whereas neural models estimate the probability of a correct next response, so the same cutoff does not represent the same level of mastery. The best-balanced threshold varied substantially across models and settings. In half of the tested settings, neural models and BKT differed by more than 0.10 in their selected thresholds, although this gap became smaller when greater priority was placed on reducing extra practice and allowing more students to advance. Stricter thresholds also did not reliably reduce performance gaps and could disproportionately restrict advancement, with stronger-prior students advancing up to 3.26 times as often as weaker-prior students. These results show that mastery thresholds should be recalibrated when the KT model or instructional priorities change and evaluated by their effects on performance, practice, advancement, and access.
Read the full abstract
Machine Learning
Nous: Learning and Certifying Memory Decisions Before Source Calibration
Agent memory systems update state decisions from reports whose reliability may be unknown. Existing analyses of source estimation do not determine when a policy can be learned or its improvement certified without identifying the reporting channel. We study these three tasks using the same observed records. For a specified hidden Markov family with continuous source uncertainty, decision learning and powered certification have quadratic sample complexity, whereas fixed-precision source estimation has quartic complexity. We characterize a sharp identified interval for policy gain under an unknown shared-background channel and derive finite-sample certificates under bounded history dependence and conditional copying. Independently trained witness regions support general history spaces, and disagreement-conditioned auditing improves power for sparse revisions. For dependent histories, prediction-count-preserving batches cancel the unknown reporting background and admit conditional certificates. A MultiWOZ 2.4 evaluation uses text-processing policies on 1,000 human-written test dialogues with simulated audits. Balanced batches retain 2.26 percentage points of the full candidate's 7.34 percentage-point mean gain and obtain more positive certificates under weak audits. These results establish task-specific information requirements and provide an auditable policy-revision framework for Nous.