Combinatorics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Sunday 8 November
A numerical semigroup is a subset of the nonnegative integers, closed under addition and with finite complement. The size of the complement is its genus. The problem of counting semigroups by the largest gap got very important through the so-called Frobenius problem, first documented in 1884. Accordingly, the largest gap is called the Frobenius number. In the last two decades, counting by the genus has become a subject of even more intense study, mostly because of the non-solved conjectures on its monotonic and super-Fibonacci growth. It is known that the number of semigroups of each genus grows asymptotically with the genus like the Fibonacci numbers and that the number of semigroups of each Frobenius number grows asymptotically in such a way that each number doubles the previous but one. We prove a formula for the number of numerical semigroups of each Frobenius number, genus, and multiplicity (first nonzero nongap), under the assumption that the multiplicity $m$ and the Frobenius number $F$ satisfy $m\geq\frac{F+1}{3}$. It has been proved that asymptotically almost all semigroups satisfy this condition. From this formula we prove multiparameter recurrence relations. They give, under the required restriction, a multiparameter exact version of the increasing behaviours previously mentioned. Indeed, we prove that the number $n_{F,m,g}$ of semigroups with Frobenius number $F$, multiplicity $m$ and genus $g$ satisfies $n_{F,m,g}=n_{F-2,m-1,g-1}+n_{F-2,m-1,g-2}$ and that the number $n_{F,m}$ of semigroups with Frobenius number $F$ and multiplicity $m$ satisfies $n_{F,m}=2n_{F-2,m-1}.$
Artificial Intelligence
InfraRSI: Harnessing Recursive Self-Improvement for Training Infrastructure
While Recursive Self-Improvement (RSI) accelerates isolated algorithm and kernel synthesis, distributed LLM training depends on tightly coupled launch settings, operator kernels, and communication schedules, where single edits often hit out-of-memory (OOM) boundaries, compiler conflicts, measurement jitter, or silent numerical drift. We present InfraRSI, the agentic harness that turns open-ended human commands into verified training optimizations via a five-stage workflow: init locks editable scope and a fixed task contract over workload, loss, and timing invariants to prevent reward hacking, doctor profiles baseline phases and calibrates cluster noise, run co-optimizes launch knobs and code patches before re-verifying top candidates against the baseline and logging findings to a cross-hardware experience bank, and report and apply audit source paths, track failed-trial costs, and commit promoted changes. Rather than stepping greedily from a single incumbent, InfraRSI initializes from diverse experience-seeded configurations and fits global execution-time and censored peak-memory models across all trials, including OOMs. Using these models, it prices per-setting memory costs, revises miscalibrated LLM priors from live runs, and pairs memory-hungry optimizations with memory savers and verified code patches. Across four single- and multi-node dense and Mixture-of-Experts (MoE) LLM workloads, InfraRSI achieves a speedup of up to $2.75\times$ and lifts mean Model FLOPs Utilization (MFU) from $17.4\%$ to $30.0\%$, outperforming existing tuners and coding agents while transferring experience across GPUs. As it improves training infrastructure efficiency, InfraRSI forms a two-level Recursive Self-Improvement loop—revising priors online within runs and distilling experience across clusters—to push foundation model scaling.
Read the full abstract
Quantum Physics
A Microscopic Model Underpinning Macroscopic QED in Active Media
We establish a microscopic foundation for macroscopic quantum electrodynamics in active media. Starting from a driven three-level system, we derive an effective two-level description with nonequilibrium stationary populations and employ a quantum central limit approach to obtain collective Langevin dynamics. This construction yields the electromagnetic susceptibility and quantum noise-current correlations from the same microscopic stationary state, providing a direct connection to the reservoir operators of conventional macroscopic quantum electrodynamics. We show that the optical response is governed by the population difference, whereas the noise correlations depend separately on the populations of the two levels. Consequently, at optical transparency, the imaginary part of the susceptibility vanishes while quantum fluctuations remain finite. We interpret the resulting noise statistics in terms of an effective reservoir occupation and a purification involving two vacuum modes, and establish a corresponding effective-temperature fluctuation–dissipation relation. Finally, we demonstrate that the residual fluctuations at transparency can induce finite heating of a nearby quantum probe.
Read the full abstract
Game Theory
The Complexity of Computing a Competitive Equilibrium in Fisher Markets with Chores
We study the complexity of computing competitive equilibria in Fisher markets with divisible chores and positive rational disutility coefficients and earning requirements. For linear disutilities, we prove that exact equilibrium computation is CLS-complete. In particular, the problem remains CLS-hard even when the earning requirements are binary in $1$ and $2^k$, where $k$ is an integer. We prove hardness by encoding the first-order conditions of a box-constrained quadratic program in equilibrium prices and earning flows, using a compact minimum-cost flow construction. For convex constant-elasticity-of-substitution (CES) disutilities, exact equilibrium prices may be irrational. We formulate the problem of finite-precision equilibrium price computation, in which we seek rational prices with a certified market clearing error at most a given rational tolerance encoded in binary. We prove that this problem is CLS-complete for every fixed rational exponent $\rho>1$. To prove this hardness, we encode a quartic polynomial through the fourth-order expansion of the CES price potential function corresponding to the Eisenberg-Gale (EG) dual program, and show that every sufficiently accurate approximate equilibrium of the constructed market lies near uniform prices. Additionally, we also identify a tractable special case of chores Fisher markets. In particular, we provide a strongly polynomial algorithm for exact equilibrium computation in linear markets whose disutility matrix has rank at most two and earning requirements are arbitrary positive rationals. For $n$ agents and $m$ chores, the algorithm uses $O(nm(n+m))$ rational arithmetic operations and comparisons, with polynomial intermediate encoding length.
Read the full abstract
Systems and Control
Concurrent Learning-Based Inertial Parameter Estimation and Control of Quadrotors
In this letter, we consider trajectory tracking control and inertial parameter estimation of a quadrotor UAV whose center of mass (CoM) does not coincide with its geometric center. Onboard payloads like an IMU, a camera, or a battery are not necessarily placed at its geometric center. Moreover, during the flight these sensors might move because of external disturbances and cause a CoM offset. Thus, these events and equipment introduce an offset and cause the inertia matrix to be non-diagonal. We develop an integral concurrent learning (ICL)-based geometric controller that estimates the mass, the CoM offset, and the full inertia matrix online, without an accelerometer or numerical derivatives, and tracks a given trajectory corrected by the offset estimated online. We prove that the estimated parameters converge to their true values under a finite excitation condition. We validate the effectiveness of the proposed algorithm using a simulation example.
Read the full abstract
Robotics
Kimodo-Policy: Learning Kinematic Action Models for Humanoid Loco-Manipulation
Humanoid loco-manipulation requires continuous coordination of high-dimensional whole-body motion to realize task objectives. While vision-language-action models and world action models transfer semantic and visual-dynamics priors to robot control, these priors do not necessarily capture the structure of robot motion. We introduce Kinematic Action Models (KAMs), a paradigm that treats whole-body motion as a direct pretraining target and uses a pretrained motion generator as the primary action-generation backbone. KAMs ground learned temporal and cross-body coordination priors through task conditioning, visual observations, and execution feedback. We present Kimodo-Policy as an initial instantiation, combining a frozen Kimodo text-to-motion backbone with lightweight ControlNet-based visual conditioning. A root-body-hand hierarchy coordinates locomotion and manipulation, while motion-history conditioning and repeated action-chunk prediction bridge motion synthesis and closed-loop control. Following large-scale loco-manipulation pretraining, the policy achieves the highest overall success among compared policies on multi-task HumanoidArena (72.14%), competitive performance on SIMPLE, and 60% mean success across eight real-world tasks, with 678M parameters (92M trainable). Ablations support the contribution of the pretrained motion backbone. Together, these results provide initial evidence for motion generation as a complementary foundation for humanoid action learning.
Read the full abstract
Analysis of PDEs
Comparability for quadratic forms with space-dependent singular measures
This paper investigates quadratic forms $\mathcal{E}^\Gamma$ generated by configurations $\Gamma$ and families of measures that are parametrized in the space variable and singular with respect to the Lebesgue measure. A first basic result is the construction of corresponding integro-differential operators in divergence form. These operators are obtained by symmetrizing the family of singular measures. The main result is the comparability of the quadratic form $\mathcal{E}^\Gamma$ with the quadratic form $\mathcal{E}^s$ which is associated with the fractional Laplacian $(-\Delta)^s$. Comparability of $\mathcal{E}^\Gamma$ and $\mathcal{E}^s$ implies Hölder regularity and a weak Harnack inequality for solutions to integro-differential equations involving these operators.
Read the full abstract
Data Structures and Algorithms
Internal Queries for Maximal Closed Substrings
Searching for repetitions inside a selected part of a text raises a subtle boundary problem: removing neighbouring characters can create new maximal answers. We study this effect for maximal closed substrings (MCSs). A string is closed if it is a single letter or has a nonempty proper prefix occurring exactly twice, as its prefix and suffix; a closed occurrence is maximal if neither available one-character extension is closed. Previous work established $O(n\log n)$ bounds on the global MCS families of a length-$n$ text and developed enumeration algorithms. Internal queries instead ask for the MCSs of a chosen substring, with maximality evaluated inside that substring. Separating boundary occurrences from interior ones gives an $O(n\log n)$-word index that reports the $c$ answers in $O(\log n+c)$ time and supports counting and longest queries in $O(\log n)$ time. Our main result reduces reporting space to $O(n)$ words, with $O((c+1)\log^{1+\varepsilon} n)$ query time and $O(n\log^{1+\varepsilon} n)$ construction time for any fixed $\varepsilon>0$. Thus, after preprocessing, reporting requires no additive scan of the queried substring, which is particularly useful for long queries with few answers. Independently, we prove that at most $3(s-1)$ MCS occurrences of any length-$s$ substring cross a fixed gap between consecutive characters, with maximality evaluated inside that substring. The same bound holds for global MCSs contained in a length-$s$ block. A family over two letters attains $\lfloor(s-1)/2\rfloor$ crossings for every $s\ge3$, showing that the linear order of this local density bound is optimal.
Read the full abstract
Computation and Language
When Should Diffusion Language Models Reconsider Their Commitments?
Masked Diffusion Models (MDMs) are well known for fast text generation via parallel decoding. However, independent parallel sampling can commit to inconsistent predictions, and less is known about how to selectively revise these early token commitments. In this work, we introduce $\underline{\mathrm{P}}$lanned $\underline{\mathrm{U}}$nmasking-and-$\underline{\mathrm{R}}$evision Joint $\underline{\mathrm{L}}$earning (PURL) for MDMs, a training-based framework that seamlessly unifies parallel generation with selective token revision. We derive a normalized joint transition whose likelihood factorizes into edit selection, unmask selection, and token prediction. These components are trained jointly on partially masked sequences constructed according to the model's own unmasking policy. During inference, all three decisions are supported by a single shared forward pass, enabling the model to decode masked tokens and revise noisy tokens. Across logic puzzles, mathematical reasoning, and code generation, PURL improves accuracy at matched unmask budgets, with relative gains up to 47.4% accuracy improvement and up to an $8\times$ reduction in scheduled decoding rounds. Beyond these gains, our analysis further reveals general principles of revisable decoding: token edits become more effective as context accumulates; early edits can steer unfinished trajectories, whereas later edits tend to remain localized; and effective edits therefore depend on both calibrating when to intervene and learning where and how to revise. Together, these suggest that effective parallel generation depends not on revising more tokens, but on making the right commitments revisable at the right time.
Read the full abstract
Rings and Algebras
Linear bound for commuting matrix algebra II
Let $\mathbb{F}$ be a field, and let $A_1,\cdots,A_m$ be pairwise commuting $n\times n$ matrices. For $m\geq3$ and $n\geq2$, we prove \[ \dim_{\mathbb{F}} \mathbb{F}[A_1,\cdots,A_m] \leq \left\lceil C_m(n-1)\right\rceil \quad\text{with}\quad C_m=\frac{\Gamma(m)}{\Gamma((m+1)/2)^2}. \] In particular, three commuting matrices generate an algebra of dimension at most $2n-2$.
Read the full abstract
Information Retrieval
Not All Branches Are Equal: Adaptive Semantic ID Construction for Generative Recommendation
Generative recommendation models represent items as sequences of discrete tokens known as semantic IDs. Existing approaches derive these identifiers using residual or hierarchical vector quantization, assuming that tokens carry independent semantic meaning across quantization levels. Revisiting this assumption, we argue that semantic IDs should behave as hierarchical paths where each token's meaning is conditioned on its prefix. This highlights a mismatch between conventional fixed-codebook quantization and the hierarchical structure implicitly learned by generative recommenders. Motivated by this insight, we propose Adaptive Silhouette Tree (AST), a top-down divisive clustering method that constructs semantic IDs as paths within a tree structure. Unlike prior methods with fixed branching factors, AST dynamically determines the number of branches at each node by maximizing the silhouette coefficient, enabling finer partitions in dense regions and compact groupings in homogeneous ones. We further introduce constrained variants that enforce minimum probability mass during splitting to improve supervision for long-tail items. Extensive experiments on five Amazon Review datasets of varying sizes and domains show that AST significantly outperforms strong baselines. These results demonstrate that aligning semantic ID construction with the inherent hierarchical structure of generative recommenders leads to more effective recommendations.
Read the full abstract
Solar and Stellar Astrophysics
Solar cycle variations of meridional flow and differential rotation in a non-kinematic dynamo model
Flux transport dynamo models have emerged as the leading paradigm for explaining the observed solar cycle. However, most of these models are kinematic and axisymmetric. We develop a non-kinematic dynamo model by specifying large-scale flows, such as the differential rotation and meridional flow. Thus, our model does not require the $\Lambda$ effect or other parameterisations of turbulent transport to generate the large-scale flows; it only solves the fluctuating flows and thermodynamics. Our model is inherently 3D by construction, although it is currently operating in axisymmetric mode because the source for poloidal field generation (Babcock-Leighton $\alpha$) is assumed to be axisymmetric. Our model reproduces solar-cycle variations in both the meridional flow and the differential rotation – specifically, the torsional oscillation in some agreement with observations. The model naturally exhibits some variability in the solar cycle, even without incorporating stochastic fluctuations, offering a new methodology for modelling the non-kinematic solar dynamo in 3D.
Read the full abstract
Computer Vision
Dropping Tokens Without Dropping Them: Causal Attribution for Diabetic Retinopathy
In this work, we investigate whether attention-based attribution maps in a vision transformer provide causal explanations for diabetic retinopathy (DR) classification. We introduce a principled decomposition of attention into two attribution maps: the contribution ($c_t$), which satisfies the completeness axiom in explainable artificial intelligence (xAI), and the pull ($p_t$), which quantifies causal explanations. We define an identity relating $c_t$ and $p_t$ that characterizes when $c_t$ fails to capture causality. We test whether each map is causally faithful on the MAPLES-DR dataset by dropping/inserting tokens, and report Necessity ($N$), Sufficiency ($Su$), and Faithfulness Correlation ($FC$). The pull $p_t$ achieves the highest causal alignment ($FC = 0.53$, $N = 0.71\%$, $Su = 3.0\%$), outperforming Attention Rollout, Grad-CAM, and AttnLRP. Occlusion retains better sufficiency ($Su = 1.6\%$) but is about 800 times slower. Furthermore, on MAPLES-DR, $p_t$ achieves $70\%$ of the average precision (AP) of HRDecoder, a state-of-the-art supervised lesion segmentation model, despite the absence of pixel-level supervision. Together, these findings show that the pull $p_t$ and the contribution $c_t$ form computationally efficient attribution maps for reliable causal interpretation in safety-critical medical diagnostics. The code and demo are available at https://github.com/pramath-haritz/dr-xai.
Read the full abstract
Mathematical Physics
Universal Critical Relations and Nonuniversal Scaling in Inverse-Square Ising Chains
For ferromagnetic Ising chains with $n^2J_n\to B>0$, the critical excess magnetization in a field inversely proportional to spin separation determines the connected correlation. Their slowly varying profile has finite logarithmic integral exactly when the ordered thermal slope is finite. For any finite nonnegative core, an exact tail gives inverse-logarithmic correlations and a square-root cusp. A relative correction $c/\log(n+e)$ with $c>1$ gives continuously varying logarithmic powers and a linear cusp. A lower bound on its disordered susceptibility excludes the usual Kosterlitz–Thouless law.
Read the full abstract
Numerical Analysis
Suboptimality of Gaussian quadrature for Freud and logistic weights in weighted Sobolev spaces
We study the worst-case error of Gaussian quadrature for integrals over the real line weighted by Freud weights $\exp(-|x|^\beta)$ with $\beta>1$, and by the logistic density $\mathrm{e}^{-x}/(1+\mathrm{e}^{-x})^2$. The integrand functions are assumed to live in the $\rho$-weighted $L^2$-based Sobolev space of smoothness $\alpha\in\mathbb{N}$. We prove that any linear quadrature using the $n$ Gaussian nodes, irrespective of the choice of quadrature weights, has a worst-case error bounded from below by a constant times $n^{-\alpha(1-1/\beta)}$ for the Freud weights, and by a constant times $(\log n)^{-\alpha}$ for the logistic weight. Our proof is based on three simple observations. First, the worst-case error of any linear quadrature is bounded from below in terms of the number of its nodes in a fixed interval. Second, the number of Gaussian nodes in such an interval is bounded from above by means of Christoffel functions. Third, Christoffel functions admit sharp lower bounds on such an interval. For the Freud weights with even integer exponents and for the logistic weight, all steps are proved in an elementary and self-contained way; for the logistic weight, we use explicit recurrence coefficients of the corresponding orthogonal polynomials. Our results provide a theoretical explanation, for integrands of finite smoothness, of the slow convergence of Gauss–Logistic quadrature observed in the literature. In contrast, for the logistic weight and for the Freud weights with even integer exponents, the Möbius-transformed trapezoidal rule attains the optimal rate $n^{-\alpha}$. This rule does not depend on the smoothness $\alpha$ and requires only point values of the integrand and the weight.
Read the full abstract
Materials Science
Overcoming speed-accuracy trade-offs with medium-sized interatomic potentials
Current machine-learned interatomic potentials offer a choice between small, fast models and large, expressive networks. This study demonstrates how an intermediate potential can approach the limit of transferable accuracy at the cost of the cheapest existing formalisms. Specifically, we model He-bearing W-Ta-Hf-Cr-V-Ti alloys with an atomic cluster expansion that learns the radial components of each polynomial as independent linear transforms and embeds numerous densities with learned exponents. While (near-)training loss appears largely a function of model size, the error on a distinct validation set is close to that of leading potentials with ten times more parameters. Taking advantage of the new structure, we develop optimized implementations that run more than an order of magnitude faster than the larger models and many times faster than smaller models with equivalent polynomials. The potential is trained on systematically enumerated small structures but generalizes to point and planar defects and can be extended to even more complex chemistries. As a demonstration, we simulate radiation-damage cascades in billion-atom systems for hundreds of thousands of timesteps.
Read the full abstract
Logic
Infinite sumsets via model theory
We prove versions of (variations of) Erdős sumset and $B+B+t$ conjectures asked by Kra, Moreira, Richter and Robertson for semigroups equipped with pairwise commuting non-algebraic definable Keisler measures. Our results use model-theoretic methods involving the existence of many coheir independent tuples in every definable relation of positive density.
Read the full abstract
Human-Computer Interaction
Dashboard Remixing: Goal-oriented Organization and Enrichment of Visualizations and Insights
Dashboard creation requires iteratively authoring multiple views and organizing them into a structured format. However, creators often struggle to step back and assess their own work for organization and refinement due to cognitive limitations. In this work, we propose a pipeline to provide computational support for self-reflection during the design process. Our pipeline automatically infers analytical objectives, discovers implicit patterns, and structures those patterns to surface coherent analytical pathways. It further enriches existing dashboard designs by generating complementary visualizations and aggregating natural language insights based on relationships across data entities (e.g., contrasting, drill-down, and relatedness) to reveal chart connections. Our rule-based pipeline provides objective, verifiable feedback, offering an external perspective while enabling creators to tell their own data stories. Four case studies and six expert interviews demonstrate how this approach helps creators reflect on organizational alternatives and surface hidden analytical opportunities in a principled way, reducing the cognitive burden while encouraging deeper analytical exploration.
Read the full abstract
Artificial Intelligence
Safeguarding Agentic AI in Radiology Workflows: A Multimodel Evaluation of Rule-based and Large Language Model-based Verification
Purpose: To estimate how often tool-using large language model (LLM) agents record clinically relevant false or unsupported information in simulated radiology documentation and how much safeguards applied before each action reduce these errors. Materials and Methods: In this simulation study (October 2026), 48 synthetic, radiologist-reviewed base scenarios covered finding communication, tumor board preparation and follow-up management. Forty-three also had a misleading-information variant (91 variants). Six LLM agents (Claude Sonnet 5.5, GPT-6.1-Sol, GLM-5.3, MiniMax M3, Gemma-4-31B, Qwen3.8-27B) completed each variant with permissions only, deterministic rules, and rules plus a locally run LLM checker (1638 episodes). The primary outcome was at least one executed high-severity violation per episode. Risk differences were estimated per agent at base-scenario level and, exploratorily, pooled across agents, with analyses of document access and blocks. Results: With permissions only, 7.3%-27.1% of episodes had a high-severity violation (pooled across agents, 19.4%). Pooled, it was 16.7% with rules (-2.8 percentage points; 95% CI: -6.1, 0.5) and 13.2% with rules plus the LLM checker (-6.2; 95% CI: -9.5, -3.0). Adding the LLM checker to rules changed it by -3.5 (95% CI: -6.3, -0.7). No difference between agents was detected (P = .43), and task success did not decrease detectably. In misleading variants, agents opened the altered document in 83% of episodes regardless of error, and 58% of erroneous statements cited it. Most blocked actions (64%-68%) would not have caused a violation and were returned for revision. Conclusion: Layered safeguards reduced but did not prevent high-severity errors. Errors after misleading information reflected misjudged rather than missed information.
Read the full abstract
Fluid Dynamics
Hydrograd v2.0: a differentiable two-dimensional shallow-water model on unstructured meshes with steady and unsteady adjoints for calibration and closure learning
We present Hydrograd v2.0, a significantly improved differentiable 2D shallow-water model written in Python with JAX. The model solves the shallow-water equations on the unstructured meshes of production models such as SRH-2D and HEC-RAS 2D with a well-balanced, positivity-preserving finite-volume scheme that handles wetting and drying. Gradients are tailored to the type of problem. Steady states are found by Newton's method with an exact sparse Jacobian obtained by graph colouring, and their gradients follow from the implicit function theorem at the cost of one linear solve. Unsteady flows are differentiated in reverse mode with checkpointing, or with an implicit time integrator whose gradient costs about two forward runs. All gradients agree with finite differences, to about $10^{-8}$ for steady flows. On the Savannah River case, the steady water surface differs from that of SRH-2D by 4.7\,mm (root mean square). Steady closure learning was 96 times faster than differentiating through time marching and reached a lower loss. An inflow hydrograph was inverted from stage records with a peak error of 2\,%. In an unsteady roughness calibration, a singular value decomposition analysis of the sensitivity matrix showed that the stage records at the gauges are 70 to 160 times more sensitive to the total resistance of the reach than to its distribution among the roughness zones, which explains why only a gauge layout with gauges in every zone recovered all five roughness values. On one GPU, a mesh of 512,000 cells takes 7.6\,s for a forward run and 51\,s for a gradient.
Read the full abstract
Cryptography and Security
PRISM: Causal Counterfactual Minimization of Agent Tool Privileges via Information Bottleneck Role Induction
This paper shifts the object of role engineering from human users to large language model (LLM) agents, and proposes PRISM to address the over-authorization that arises when an agent obtains full tool privileges through the Model Context Protocol (MCP). PRISM formulates whether a tool is necessary for a task as an intervenable causal proposition, and applies structured interventions in a sandbox to learn a counterfactual necessity model. It then compresses trajectories into discrete latent variables through an information bottleneck, and determines the number of roles from the knee of the rate-distortion curve. Finally, it formalizes pruning as a bilevel Stackelberg game, in which a generative flow network samples diverse near-optimal solutions. On three public benchmarks, AgentDojo, $\tau$-bench and ToolEmu, PRISM compresses the number of authorized tools per role by 78.3%, attains a task success rate of 78.2%, reaches a counterfactual necessity accuracy of 88.7%, and lowers the blast radius to 15.9%.
Read the full abstract
Pattern Formation and Solitons
Axis-switching oscillations of elliptic scroll rings in excitable media
Scroll waves in three-dimensional excitable media rotate around a filament that moves according to its local curvature. Using simulations of the Barkley model, we show that initially planar, elliptic scroll rings perform the axis-switching oscillation known from elliptic vortex rings in fluids: while the ring drifts along its axis, it buckles into a saddle, becomes planar again with interchanged axes, and repeats this cycle. The oscillation occurs for vanishing, positive, and negative filament tension. As predicted by a kinematic filament equation, its period scales approximately with the square of the ring radius and inversely with the drift coefficient. For our main model parameters, however, the oscillation is about 1.45 times slower than the linear prediction based on the drift of circular rings of the same size. A curvature-dependent drift coefficient accounts for a factor of about 1.25, while the remaining slowdown results mainly from a weaker coupling of out-of-plane to in-plane deformations, which a fourth-order rigidity term does not describe. In contrast, the Panfilov-Rudenko model and the Atri model of Ca$^{2+}$ waves follow the linear prediction within about 6%, showing that the slowdown depends on the kinetics.
Read the full abstract
Quantum Physics
Weight-Space Programming of Variational Quantum Circuits: Large Language Models Write Quantum Circuit Parameters Without an Optimizer
Variational quantum algorithms (VQAs) optimize circuit angles through repeated quantum-device evaluations, often facing vanishing gradients and high optimization costs. We ask whether a large language model (LLM) can write these angles directly. Extending *weight-space programming* from classical networks to quantum circuits, we provide a frontier LLM (\gpt) with the problem, circuit, and gate conventions. Its angles are executed exactly as written, without returning losses, gradients, or accuracy feedback. Across 139 requests, we test handwritten-digit classification, VQE ground-state estimation for spin chains and H$_2$, and MaxCut QAOA, comparing against random weights, classical optimization, and optimization initialized with LLM weights on the same circuit. When parameters have interpretable structure, the LLM matches or approaches the optimizer: it achieves 95–100% of the energy spectrum on every VQE problem, derives the exact H$_2$ ground state in closed form, finds optimal depth-one QAOA angles for every graph, and reaches 93–96% accuracy on digits 3 versus 8 with 12–18 hand-written weights, outperforming an optimizer trained on the same ten examples. Its weights are also strong warm starts: optimization reaches its best value after a median of one evaluation rather than hundreds. Allowed to design circuits, the LLM chooses data-dependent rotation chains encoding pixel importance in all 30 cases (one qubit for binary tasks, one chain per qubit for ten classes), reaching 65% accuracy across ten digits. Yet the approach fails in opaque weight spaces: accuracy drops to chance when fixed circuits exceed 30 weights, while optimization improves. It is also sensitive to gate conventions: 6 of 42 QAOA replies reverse mixing-angle signs and fail. We release all prompts, replies, and scores.
Read the full abstract
Functional Analysis
On Słociński-type decomposition, BCL model and unitary extension of commuting Banach space Wold isometries
The adjoint of an operator plays the central role in Hilbert space operator theory. In particular, dilations of Hilbert space contractions, extension of commuting Hilbert space isometries to Hilbert space unitaries, orthogonal decompositions of contractions and functional models for commuting isometries are heavily dependent on the presence of adjoint. In this paper, we develop a theory of Banach space Wold isometries with right-complemented ranges by showing explicit constructions of Berger-Coburn-Lebow model and unitary extension for such isometric tuple without the help of adjoint. We also show that a Słociński type decomposition is possible for such a commuting tuple of Banach space isometries. Each of our results generalizes the corresponding result in the Hilbert space theory.
Read the full abstract
Logic in Computer Science
Addressing Bias Early in the AI Lifecycle: Social Determinants of Health in Functional and Cognitive Assessment Labels of Nursing Home Residents
Fairness in clinical artificial intelligence (AI) is usually evaluated after model training, yet bias can enter earlier in the AI lifecycle, during data collection and labelling. When routine clinical assessments serve as target variables, systematic differences in these labels across sociodemographic groups are carried into model development, validation and updating. This study addresses this early lifecycle stage by auditing labels from three standardised instruments used in residential care: the Barthel Index (BI), the Mini-Mental State Examination (MMSE) and the Global Deterioration Scale (GDS). Using electronic health records of 2,407 residents from 12 nursing homes in the Basque Country, Spain (2011-2021), we assessed four social determinants of health (SDoH) - sex, educational attainment, lifetime work activity and marital status - with two group fairness metrics, Disparate Impact (DI) and Statistical Parity Difference (SPD), computed for each outcome category with bootstrap confidence intervals. Disparities were found for all four attributes and across the three instruments: men, residents with intermediate education and never-partnered residents were more frequently classified in categories indicating preserved function or cognition. These disparities may reflect genuine differences in health status, measurement properties of the instruments, assessment practices or selective admission to residential care, and cannot be attributed to a single cause with the present design. Regardless of their origin, any model trained on these labels would inherit them, and they could persist through later lifecycle stages unless explicitly examined. These findings support pre-training label auditing as the first step of a continuous fairness assessment spanning model development, validation, deployment and updating.
Read the full abstract
Mathematical Physics
SU(2) Yang-Mills: Spectral Bounds, Vacuum Concentration, and Criteria for the Continuum Limit
A continuum mass-gap argument must preserve nonzero renormalized states and a positive spectral estimate while the spatial cutoff is removed and physical volume grows. We develop quantitative results that separate these requirements. Gauge reduction, reflection positivity and attributed strong-coupling estimates give physical-transfer bounds with nontrivial local sectors. At fixed spatial cutoff and positive coupling, we construct subsequential infinite-volume physical-time dynamics from local history correlations, with nonconstant gauge-invariant observables. For finite spatial boxes, explicit bounds apply to the entire physical vacuum complement; finite-matrix criteria include all omitted representations. Vacuum concentration gives extensive entropy bounds, quantitative fidelity obstructions for hard electric cutoffs, and necessary growth of retained representation spaces. A sharper local comparison yields complete bounds on fixed discarded-link sectors, while collective projection and Schur estimates isolate the additional input needed for a volume-uniform gap. The work also certifies a finite three-link conditioning functional through Wilson coupling 0.354, with its physical interpretation conditional on the stated block-comparison hypotheses. Exact blocking, corrected effective-potential theorems and a representation-by-representation two-dimensional benchmark clarify which approximations preserve the relevant observables. Published glueball measurements retain their original attribution; their reanalysis and the reproduced winding-channel and replica studies have explicitly statistical scope. Complete derivations and reproducible calculations support a set of certified benchmarks and continuum transfer criteria. A nontrivial four-dimensional spatial-cutoff construction and a positive spectral bound along it are not established; no exact continuum mass is asserted.
Read the full abstract
Distributed Computing
Ostinato: Fast CPU-GPU MoE Inference for Local Agents via Resident Experts in Every Forward Pass
Language-model agents increasingly run on workstations with a few consumer GPUs, where large mixture-of-experts (MoE) models keep part of their experts in host memory. With existing CPU-GPU engines, these agents get fast prefill or fast decode, not both: llama.cpp's expert cache skips forward passes of more than 32 tokens to keep the experts of decode cached, so an agent waits 2.1-2.8 s after each tool output, whereas its default layer placement shortens this wait but is 1.5-1.6$\times$ slower in decode. Our key insight is that a short prefill touches too many distinct experts to upload, yet sends 84% of its expert activations (routes) to a stable set of experts that fits in GPU memory. We present Ostinato, a CPU-GPU MoE inference engine that computes these experts on the GPUs in every forward pass without changing what the model computes. Short prefills split their routes, leaving fewer than one of a token's six routed experts per layer to the CPU; long prefills upload each distinct cold expert once; and each decode step runs as one fixed-shape CUDA graph per GPU. On real coding and customer-service agent sessions with DeepSeek-V4-Flash on four RTX 4090D GPUs, Ostinato is the fastest engine in both phases and finishes every session sooner than all six baselines. Against llama.cpp and KTransformers, it cuts the median wait after a tool output by 22-55% and finishes sessions 1.4-2.5$\times$ sooner.
Read the full abstract
Artificial Intelligence
DriveFoundry: Co-Evolving Data and Model for Efficient Closed-Loop Iteration in Autonomous Driving
A driving foundation model must learn reusable decision knowledge from rare critical events and apply it under strict runtime constraints. We introduce DriveFoundry, a closed-loop framework improving iteration efficiency under data and computation budgets by coupling structured causal data refinement with adaptive inference. The Data Engine builds traceable scene-candidate records linking scene evidence and factor-decision relations to candidate trajectories and outcome assessments. Candidate comparisons and counterfactual annotations clarify how scene factors affect behavioral feasibility and preference. Verified model failures guide VLM-assisted annotation and expert correction, refining training data and annotation systems. The model engine, Xiaomi-Drive-o1, unifies Perception, Causation, Action, World, and Value within a shared vision-language backbone to enable cross-task knowledge transfer. Block-wise latent refill passes candidate-specific evidence and predicted visual consequences between task blocks, connecting scene understanding, planning, and evaluation. A learned value-of-computation head selects tasks and allocates latent tokens based on expected benefit and cost, adapting inference to each decision. On PhysicalAI, the 9B model achieves the lowest ADE@1 across evaluated splits and horizons among tested models. In AlpaSim, it outperforms Alpamayo 1.5 across all metrics and achieves lower close-encounter and off-road rates than the 34B Alpamayo 2 Super, while trailing in progress and AlpaSim scores. For the 0.8B variant, joint training reduces 5 s minADE@6 by about 54% without altering the inference path or token budget. Adaptive inference matches full reasoning performance using only 28% of tokens. Real-vehicle tests confirm runtime feasibility, supporting a unified paradigm for learning from critical experience and dynamically allocating computation.
Read the full abstract
Machine Learning
Probing Transformer Representations Layer by Layer, Module by Module
Which representation should be probed in a transformer-based foundation model? Recent work shows that the standard practice of probing the final layer is often suboptimal, as intermediate layers can provide better features. Yet, representations are still largely studied along depth alone, treating each transformer block as an atomic unit. In this work, we analyze transformer representations layer by layer, module by module. Through more than 144,000 linear probing evaluations of vision foundation models (VFMs) and large language models (LLMs), we show that layer and module choices should be considered jointly: the best layer-module pair yields substantially larger gains over the final-layer residual stream than the best layer with the residual stream fixed. Moreover, module performance can vary substantially across layers, showing that useful representations cannot be characterized by depth alone. Building on these findings, we evaluate simple fixed probing strategies: the final-layer feedforward activation (Act@100) for vision and the second normalization layer at 60% depth (LN2@60) for language. These strategies improve over the probing baseline in a majority of evaluations and recover, on average, 42% and 86% of the gains from probing the best layer-module pair for vision and language, respectively. For VFMs, these gains increase with the transfer difficulty between pretraining and downstream data.
Read the full abstract
Analysis of PDEs
Asymptotic Stability of Planar Viscous Contact Waves for the Three-Dimensional Full Compressible Navier–Stokes Equations in the Whole Space
The asymptotic stability of planar viscous contact waves is established for the three-dimensional full compressible Navier–Stokes equations in the whole space. The exact evolving planar background is constructed in a companion one-dimensional paper and may arise from a large integrable perturbation. The three-dimensional remainder has small energy, prescribed second-derivative bounds for density and velocity, and a prescribed first-derivative bound for temperature. Its initial temperature lies in a fixed positive interval; its second derivatives need only be finite. A unique global positive strong solution and algebraic decay in the uniform norm are obtained. The wave-strength threshold depends on lower-order planar bounds, while the allowed remainder size may depend on the complete planar datum.