Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Friday 6 November

Machine Learning

Projected Neural Additive Models as Universal Approximators

Neural additive models (NAMs) offer interpretability but lack expressivity due to their rigid additive structure, whereas multi-layer perceptrons (MLPs) achieve universal approximation at the cost of transparency. We establish the first theoretical foundation for projected neural additive models (PNAMs), which augment NAMs with a learnable linear transformation $\boldsymbol{T}$ to capture cross-variable interactions. While standard NAMs cannot approximate even the simplest interaction term $\chi_1 \chi_2$ because their cross-derivatives are identically zero, we prove that PNAMs are universal approximators and characterize the minimum number of projections required for exact polynomial representation. Furthermore, we derive an approximation rate of $\mathcal{O}(M^{-s / (N - 1)})$ for $\mathcal{C}^s$ functions (i.e., functions with $s$ continuous derivatives), where $M$ denotes the number of projections and $N$ the input dimension; this rate is known to be optimal for ridge function approximation. To recover interpretability from the projected inputs, we introduce structured sparsity regularizers and post-hoc symbolic regression techniques that enable global input ranking, parameter pruning, and the conversion of learned bases into compact analytical expressions. Experiments on knot theory invariants, MNIST, and phase field fracture mechanics demonstrate that PNAMs match or surpass similar architectures in accuracy while discovering salient features and yielding symbolic models with orders of magnitude fewer parameters.
Read the full abstract

All authors Nhon N. Phan, Qiang Du, John D. Clayton, WaiChing Sun

Cryptography and Security

GSynPEF: A Privacy Evaluation Framework for Synthetic Graphs

It is often desirable to generate and release synthetic graphs, which resemble real-world data, in a privacy-preserving way. While substantial progress has been made in designing private graph synthesizers (PGS), opportunities remain for more explicit connections between synthesizer models and privacy evaluation, and better accounting for output variation. This paper proposes GSynPEF, a practical framework for empirical evaluation of privacy and utility in settings for which a real-world graph is given as input and output synthetic graph(s) are observed. Our framework links a broad class of graph synthesizers (both statistical and deep-learning based) with privacy metrics, expanding evaluation dimensions to practical utility and risk, through empirical analysis of multiple actual outputs. The framework accommodates both an Exact Solution scenario, where synthesizer behavior is directly observable, and a Monte Carlo Approximation scenario where only synthetic outputs from a black box synthesizer are available. Extensive experiments on four real-world datasets and seven widely used graph synthesizers demonstrate framework's versatility, enabling privacy evaluation for different attack goals (e.g., to infer an edge or node) and capabilities (e.g., observing one or multiple synthetic graphs). We analyze evaluation results by examining the relationship between graph statistics and synthesizer design choices, providing insights for data curators in PGS selection and researchers for improved PGS design.
Read the full abstract

All authors Yu Duan, Luca Baldesi, Carter T. Butts, Athina Markopoulou

Data Structures and Algorithms

Fully Dynamic Correlation Clustering with Vertex Updates in Polylogarithmic Update Time

We present a fully dynamic algorithm for correlation clustering that supports vertex insertions and deletions while maintaining an $O(1)$ approximation in polylogarithmic amortized update time. This resolves an open problem of vertex deletions posed by a recent line of work on dynamic correlation clustering. We adopt the min-disagreement objective for correlation clustering, i.e., given a signed complete graph $G=(V,E^+\cup E^-)$, we want to minimize the total number of $(+)$ edge crossing clusters and $(-)$ edges within clusters. The input is given in the adjacency list model, where each update inserts or deletes a node along with its full set of $(+)$ neighbors. Our main technical contribution is a fault-tolerant sparse-dense decomposition framework which unlike the existing decomposition-based frameworks allows for structural inconsistencies as long as the induced cost is bounded. We also show a separation between explicitly maintaining the (stronger) sparse-dense decomposition and maintaining a constant-factor approximation to correlation clustering, which may be of independent interest.
Read the full abstract

All authors Vladimir Braverman, Prathamesh Dharangutte, Chenglin Fan, Chen Wang, Shenghao Xie, Samson Zhou

Superconductivity

Current-voltage characteristics of layered cuprate superconductors with intrinsic Josephson junctions

A brief review of the results for the current-voltage characteristics study in highly anisotropic cuprate superconductors is presented, considering the existence of the intrinsic Josephson effect in multilayer HTSC compounds. We analyzed the observations of multi-branch current-voltage characteristics for current in the c-direction both for the hole-doped and electron-doped cuprate SCs wich conveniently viewed as a strack of intrinsic Josephson junctions, where the CuO2 layers serve as superconducting plates and the buffer layers as dielectric interlayers. It was shown that the tunneling characteristics of intrinsic Josephson junctions in optimally doped p- and n-type cuprates indicate a predominantly d-wave symmetry of the order parameter, demonstrating high conductivity in the subgap regime: U < 2delta/e. The analysis carried out confirms the theoretically predicted power-law dependences U ~ I^gamma with gamma = (0.33 - 0.50) in the hole-doped Bi- and Tl-compounds, which is typical for SCs with monotonic dx^2-y^2 - wave symmetry of the order parameter and with gamma = (0.70 - 0.76) for electron-doped SCs with a non-monotonic d-wave symmetry within the framework of the two-band/two-gap model.
Read the full abstract

All authors T.B. Charikova, V.N. Neverov, M.R. Popov, S.D. Popov, N.G. Shelushinina

Computer Vision

ATLAS: A Multi-Procedure Benchmark and Context-Aware Model for Surgical Anatomy Segmentation

Recognizing anatomical structures during minimally invasive surgery (MIS) could support intraoperative navigation and improve patient safety, but progress is limited by small, single-procedure datasets and by segmentation methods developed for natural scenes. This work addresses both limitations through a new dataset and model. First, we introduce ATLAS-120k, a clip-level semantic segmentation dataset of 121,018 annotated frames from 100 surgical videos, covering 14 laparoscopic and robot-assisted procedures and 41 classes. Annotations were created with a pipeline that combines manual labeling, tracker-based propagation, and review by surgeons. Second, we propose ATLAS (Anatomy Recognition with Context Learning using Foundation Representations), a video semantic segmentation model for surgical anatomy recognition. ATLAS builds on in-domain foundation-model embeddings and adds context queries for the procedure and surgical phase, together with query propagation between frames for short-term memory. ATLAS outperforms state-of-the-art natural-image and in-domain foundation models in both segmentation accuracy and temporal consistency, achieving a Dice score of 0.70 with ViT-L while running in real time at 76.2 FPS. Ablations show that the context queries improve accuracy and that query propagation improves temporal consistency. ATLAS also generalizes well, matching a dedicated model zero-shot on CholecSeg8k and improving phase recognition on AutoLaparo over SurgeNet initialization. The model weights, dataset annotations, and annotation tool are publicly available at https://github.com/TimJaspers0801/ATLAS.
Read the full abstract

All authors Ronald L. P. D. de Jong, Tim J. M. Jaspers, Raf A. H. Vervoort, Aron F. H. A. Bakker, Anna L. Dijk, Yiping Li, Jip L. Tolenaar, Jelle P. Ruurda, Willem M. Brinkman, Daan de Geus, Fons van der Sommen, Josien P. W. Pluim, Marcel Breeuwer

Data Structures and Algorithms

A Note on the Hardness of Partially Dynamic Max Flow

We study the fine-grained complexity of partially dynamic max flow in capacitated graphs. Assuming the combinatorial 4-clique hypothesis, combinatorial algorithms require $\Omega(n^{2-o(1)})$ amortized update time. Thus, recomputing max flow after each update is optimal up to subpolynomial factors for combinatorial algorithms in dense graphs. The previous conditional lower bound was $\Omega(n^{1-o(1)})$ amortized update time under SETH or the OMv hypothesis and applies to all algorithms [Dah16].
Read the full abstract

Artificial Intelligence

Generative Active Learning for Label-Efficient Pan-Organ Ultrasound Diagnosis

Medical ultrasound (US) enables real-time, non-invasive diagnosis, yet building US-based AI systems is limited by data scarcity, costly expert annotation, and limited cross-center generalizability. Privacy constraints and instance-level reporting hinder the collection of diverse abnormal cases and frame-level labels, while repeated retraining for new centers increases computational cost. We introduce EvoDiff, a generative active learning framework that integrates diffusion modelling with information-theoretic objectives, evolutionary search, and curriculum-based sampling to select informative samples across heterogeneous US datasets. EvoDiff formulates sample selection as a generate-to-retrieve process, in which diffusion-based optimization in feature space guided by information gain produces synthetic probes, and only the real unlabeled samples retrieved by these probes are annotated and used for downstream training. Across six benchmarks covering fifteen datasets and 16,696 images, we evaluate EvoDiff on convolutional network and transformer-based foundation model. EvoDiff approaches fully supervised performance with substantially fewer labels, outperforms active learning baselines, and improves foundation model generalization under cross-domain external validation on unseen cohorts. A reader study with six radiologists indicates the potential to support clinical annotation workflows, with a model-based estimated 24% to 84% reduction in annotation workload at matched downstream performance.
Read the full abstract

All authors Yaofei Duan, Yuhao Huang, Yuan Gao, Ruixiang Qi, Tianyu Zhang, Yanlin Geng, Yue Sun, Sio-Kei Im, Dong Ni, Dong Xu, Ritse Mann, Tao Tan

Systems and Control

Open-Source Multi-Fidelity EMT Modeling of Data Center Loads for Power System Studies

Characterizing data center-grid interactions under rapid load fluctuations requires electromagnetic transient (EMT) models that capture fast converter and control dynamics. This paper develops a modular multi-fidelity EMT model of data center loads in PSCAD. The modular design supports flexible assembly of subsystem models for different power supply configurations. The multi-fidelity design incorporates switching models, average-converter models, and reduced-order models to balance simulation accuracy and computational efficiency according to analysis needs. The model also provides a systematic controller tuning procedure and supports control mode switching. Comparative simulations demonstrate how this data center load model responds to both grid-side and load-side disturbances and assess the accuracy and computational efficiency of models with three fidelity levels. The model is publicly available on GitHub.
Read the full abstract

Operating Systems

The Storage View Of KV Cache: Characterizing Cache Behavior Across LLM Workloads

Modern LLM serving increasingly extends KV caches beyond GPU memory into host DRAM, SSDs, and remote storage, yet these stores remain largely governed by workload-agnostic policies such as LRU and fixed TTLs. We ask what execution context is needed by storage to make retention, reclamation, and placement decisions across heterogeneous workloads. We characterize KV-cache reuse, lifetime, and footprint at block granularity across 11 production and public traces spanning chat, API serving, RAG, coding agents, tool-using agents, and multi-lineage workflows. We examine reuse sources, recurrence timing and frequency, useful lifetimes, and storage consumption. Across workloads, KV state takes two recurring forms. Single-session state is reused within one execution: its footprint grows as sessions deepen and accumulates across concurrent sessions; it can remain useful through long idle periods but loses value at session completion. Multi-session state recurs across sessions of the same workload, outlives individual sessions, and spans hot prefixes and a large infrequent tail. While prior systems often exploit application-, agent-, or workflow-specific knowledge, our results suggest that three generic signals (a workload identifier, a session identifier, and session completion) expose much of the structure needed for KV-storage management. Trace-driven comparisons show that workload-specific recurrence estimates retain up to 22% more within-session hits and reduce dead state by up to 40%; a session completion signal eliminates the fundamental idle-time reclamation tradeoff for single-session state; and session-aware reuse classification identifies shared state 4-700x more selectively than request-level access counts. These results motivate a small cross-layer interface for workload-aware KV storage without application-specific semantics.
Read the full abstract

All authors Talia Q. Chen, Daniel Waddington, Narangerelt Batsoyol, Vijay Chidambaram

Mesoscale and Nanoscale Physics

Optimizing one-dimensional lithographically defined nanowires in semiconducting heterostructures

One-dimensional nanowires can host remarkable electronic properties and are central to studying quantum transport. Although heterostructure-defined confinement combined with gate-defined potentials can greatly reduce the effects of impurities as measured via electron mobilities, the realization and interpretation of novel effects at this scale are complicated by disorder. Experiments on lithographically defined one-dimensional nanowires therefore do not exhibit ideal behavior. Using theoretical and computational models, we predict that optimizing the geometry of split gates in semiconducting heterostructures will lead to higher-quality channels for transport measurements. Further, we show that this optimization is essential for resolving features of interest such as the Rashba-induced spin gap. This new approach to lithographically defined conduction channels provides better energy resolution and is less susceptible to disorder.
Read the full abstract

All authors Leah Tom, Yik Kheng Lee, Karina L. Hudson, Krittika Kumar, Jackson S. Smith, Alex R. Hamilton, S. N. Coppersmith, Jared H. Cole, Mark Friesen

Artificial Intelligence

AVCORP: Identifying and Resisting Deception via Communication Theory and Theory of Mind

Today's LLM-based agents are vulnerable to deception by users and other agents. Defending against deception requires Theory of Mind (ToM) reasoning to infer the beliefs and goals of others as well as modeling tactics based on communication theory. To enable training models to detect deception, we adopt the well-known deception game, Avalon. We present a corpus, AVCORP, of 5,000 annotated utterances from 250 games of Avalon. AVCORP includes (1) 37 tactics mapped to a 4x4 behavior matrix, and (2) reasoning traces capturing abductive hypotheses, suspicion levels, and nested ToM beliefs. We train Qwen3-8B on AVCORP and evaluate it on 25 held-out games and 20 games played by humans. On held-out games, our best trained detector outperforms its untrained baseline and frontier models (GPT-5.6 Sol, Gemini 3.1 Pro, DeepSeek-V4-Pro), matching Claude Opus 5.5. On human games, every trained detector exceeds the untrained model in F1 score across all rounds.
Read the full abstract

All authors Fardin Saad, Zhe Zhang, Sanjana Cheerla, Amanul Haque, Munindar P. Singh

Computation and Language

ASCENT: Scale-Complementary External State for Frozen Long-Context Language Models

External state should become more valuable as a reader becomes more capable, yet long-context evaluations measure memory utility at a single model scale. This paper formulates scale complementarity: the absolute gain from external state increases when reader capacity and a compact state budget are jointly scaled. ASCENT realizes this principle with a target-blind causal writer, nested typed state, and a bounded query-conditioned readout; every language model remains frozen. Under log loss, the value of a nested refinement equals its target-relevant conditional information; a full factorial makes the reader-by-state interaction identifiable. Official 16K BABILong experiments show Qwen2.5 accuracy gains of 10.50, 78.88, and 82.63 percentage points at 0.5B, 1.5B, and 3B parameters, with both adjacent directional contrasts significant after family-wise correction. At 7B, Foundation accuracy rises to 33.25% while four-slot ASCENT answers every question, which caps the gain at 66.75 points; the 3B-to-7B accuracy increment is -15.88 points [-20.55, -11.20]. The accuracy-side claim is therefore established from 0.5B to 3B; the 7B conclusion rests on the non-saturated negative-log-likelihood factorial. A three-model by three-state factorial yields co-scaled negative log-likelihood gains of 0.388, 2.099, and 7.406 nats at 1.5B, 3B, and 7B; all interaction and gain contrasts retain positive simultaneous lower bounds. Five public 7B-8B readers obtain 66.75-85.63 point gains, and ASCENT exceeds task-trained retrieval checkpoints at the two larger shared scales. Controls, baselines, systems accounting, matched-hardware reproduction, and a cross-architecture sensitivity analysis accompany the experiments. The results establish scale-complementary external state under the evaluated benchmarks, metrics, and state schema.
Read the full abstract

Computer Vision

CXR-CMFA: A Cross-Modal Feature Aggregation Network for Chest X-Ray Pathology Classification

Chest X-ray (CXR) interpretation remains challenging because subtle pathology can be obscured by overlapping anatomy, image-quality variation, and limited visual evidence in isolated radiographs. Although image-only deep learning models provide useful baselines, they discard the complementary diagnostic context available in radiology reports, while large vision-language models can be computationally expensive for practical deployment. This paper presents CXR-CMFA, a lightweight cross-modal feature aggregation network for binary CXR pathology classification. The proposed framework combines a dual-stream visual encoder, using EfficientNet-B0 for global anatomical representation and DenseNet-121 for local pathological detail, with a TF-IDF-MLP report encoder. Two fusion modules integrate these modalities: the Dual-Modal Spatial Attention Module (DMSAM) uses the report embedding to gate global and local visual features, and the Cross-Modal Gated Feature Fusion (CGFF) module learns per-instance weights for visual and textual contributions. Experiments on the MIMIC-CXR benchmark, using the provided train, validation, and test splits, show that CXR-CMFA achieves 93.45% accuracy, 99.88% precision, 91.8% recall, 95.7% F1-score, and 99.19% AUC on the held-out test set. The model produces only four false positives among 984 normal test cases, supporting its suitability for high-specificity screening workflows. An ablation study further demonstrates a 22.47 percentage-point AUC improvement over the global-image-only variant, confirming the value of adaptive cross-modal fusion. These results indicate that compact report-conditioned fusion can provide strong discriminative performance without relying on large pretrained language models.
Read the full abstract

All authors Talukder Hasnat Zadid, Jahin Sultana, Felipe Bochini Brisola, Kishore Roy, Mysha Rahman, Syeda Asrafa Islam, Md. Abdullah-Al-Jubair

Materials Science

Hidden memory above the Neel temperature programs antiferromagnetic order in Cr2O3

Control of antiferromagnetic order is central to magnetoelectric and antiferromagnetic spintronic devices. In the prototypical magnetoelectric antiferromagnet Cr2O3, domains are conventionally selected by simultaneous application of electric and magnetic fields during cooling through the Neel temperature (T_N) or at a fixed temperature below T_N. Here, we show that antiferromagnetic order can instead be programmed by electric-field conditioning entirely in the paramagnetic state above T_N. By measuring the magnetoelectric coefficient (alpha_zz) of bulk Cr2O3 crystals under different thermal and field histories, we find that a sufficiently large electric field above T_N establishes a reversible state that persists after the field is removed. Subsequent magnetic-field cooling with no applied electric field produces a single-domain state with the opposite Neel-vector orientation from conventional magnetoelectric field cooling. The electrically induced state can be reversibly written, erased, and allowed to relax at high temperatures, demonstrating a long-lived internal electric bias that governs domain selection. The observed behavior is consistent with a metastable surface-associated polar state that may generate this internal bias. These results reveal an unexpected coupling between electrically programmed memory and antiferromagnetic domain formation in Cr2O3.
Read the full abstract

All authors Josiah Keagy, Sumukh Mahesh, Wei-Cheng Liao, Jing Shi

Optimization and Control

A Hierarchical Variational Inequality Approach to Generalized Nash Equilibrium Seeking with Nonlinear Shared Constraints

We study the computation of variational generalized Nash equilibria (v-GNE) in games with strongly monotone pseudo-gradients and nonlinear convex shared constraints. We show that the v-GNE problem is exactly equivalent to a hierarchical variational inequality, whose lower level is given by the gradient of a smooth convex penalty of the constraint violation over the local action sets. Building on this reformulation, we propose a hierarchical regularization scheme that operates entirely in the primal space and requires only projections onto the local action sets. In the centralized setting, we prove last-iterate squared distance convergence to the v-GNE with rate $O(1/t^{\scriptscriptstyle 2/3})$ under a local error bound condition, which holds, e.g., under Slater's condition, improved to $O(1/t)$ under a global error bound condition. We then extend the scheme to the partial-decision information setting, where the agents communicate over a network and maintain local estimates of the joint action. We derive a distributed algorithm with a constant stepsize operating in the primal space and converging to the v-GNE with rate $O(1/t)$.
Read the full abstract

Optics

How light interacts with matter, revisited

In 1968 Victor Weisskopf explained in Scientific American how light interacts with matter, with a few formulas and many words. We tell the same story again with the tools that have since become available: the fields are computed and drawn, and their evolution in time is shown in animations. Light is described as a classical electromagnetic field and matter as a collection of charged oscillators; every phenomenon follows from adding up the retarded waves radiated by the oscillators, each with its proper delay and phase. From this single picture we obtain the blue of the sky and the white of the clouds, transparency and the refractive index, the Ewald-Oseen extinction theorem, reflection and the colours of thin films, why metals reflect and what gives metals and semiconductors their colours, polarizers, Brewster's angle and birefringence, the working of mirrors, lenses, microscopes and the eye, the propagation of a few-cycle pulse through a slab, Cherenkov radiation, amplification and second-harmonic generation. Feynman's arrows are made quantitative with the sheet method, and coupled-dipole simulations show the incident wave being cancelled and replaced inside the medium. The connection with quantum theory is discussed where the classical picture stops: the photon, spontaneous emission and the forbidden lines of the aurora, vision and photosynthesis, and the sum over paths. The text is addressed to undergraduate students and their teachers, and to any reader curious about the physics behind everyday optical phenomena; derivations are collected in separate panels, and the animations and the Python codes used for the figures and simulations are provided as ancillary files.
Read the full abstract

Computation and Language

Where Did the Evidence Go? Diagnosing Proof-Set Loss in Long-Term Conversational Memory

Long-term conversational question answering can require several separated dialogue turns that support an answer only together. We introduce EviCover, a retrieval approach designed to retain this complementary evidence. It converts questions into observable evidence requirements, checks requirement coverage using retrieved source identifiers, and selects turns jointly. We track annotated support sets - the benchmark-labeled dialogue turns associated with each question - through retrieval, reranking, and final answer-generation context. We evaluate the baseline memory pipeline on generated, human-verified LoCoMo conversations and human-written REALTALK conversations, and examine EviCover on REALTALK's 108 multi-hop questions. Compared with the hybrid lexical/dense baseline, support-turn recall among 60 candidate turns rises from 0.526 to 0.575, and recall in the final ten-turn context rises from 0.349 to 0.379. The evidence funnel identifies separate discovery and selection failures, while paired answer evaluation shows that greater evidence recall does not yet establish a correctness gain. This analysis locates where complementary evidence is lost and distinguishes evidence retrieval from its use in answer generation.
Read the full abstract

Robotics

Capturability-Aware Footstep Planning in Cluttered Environments: A Recursively Feasible Sequential MPC on Axis-Aligned Safe Corridors

A fundamental aspect of humanoid footstep planning is to jointly guarantee obstacle avoidance and locomotion stability. This paper presents a capturability-aware sequential Model Predictive Control (MPC) framework for Linear Inverted Pendulum (LIP)-based humanoid footstep planning in the presence of obstacles. The proposed method combines obstacle avoidance with capturability constraints through safe corridors constructed as sequences of intersecting rectangles, yielding a sequential MPC framework with recursive feasibility guarantees for the LIP-based planning layer. The planner autonomously adapts to constrained scenarios, switching to in-place walking when obstructions—such as doors assumed to be open during planning but unexpectedly found to be closed during execution—block forward motion. Performance is largely independent of the prediction horizon, allowing high success rates while keeping the computational cost low. The approach is validated in high-fidelity simulations with the humanoid robot Digit, demonstrating a practical solution for capturable humanoid footstep planning in cluttered environments.
Read the full abstract

All authors Athanasios Kyriakopoulos, Shubham S. Kumbhar, Kunal S. Narkhede, Abhijeet M. Kulkarni, Ioannis Poulakakis

Artificial Intelligence

Key Institutional Considerations for Designing Effective Independent Verification Organizations

Independent verification organizations (IVOs) have emerged as a popular, bipartisan solution for facilitating high-quality independent evaluations of AI technologies. But momentum is outpacing design: there is emerging consensus on the necessary operating conditions for robust IVOs, but little discourse on institutional and policy specifications to ensure those conditions are met. IVOs are currently envisioned primarily as a mechanism operating at the AI frontier and model level to evaluate catastrophic risks. Little work has yet been done on how to establish or enforce an IVO ecosystem that is equipped to handle myriad risks (i.e., beyond catastrophic ones) and operate at the system and application level across the deployment and use stages of the AI lifecycle. This paper argues that IVO design should ensure the capacity for evaluating AI models, systems, and applications in context for both frontier companies and enterprise companies & AI-adopting organizations. Additionally, this paper provides early considerations to advance IVO discourse from normative to operative, and from existential to everyday risks. Drawing from existing institutional constructs in financial services, education, public health, and science, we outline key considerations on whom IVOs serve, what access they require, how they interact with external governing structures, how they are funded, and how they fit into the wider assurance ecosystem. Much of the structure and knowledge necessary to holistically design IVOs and their ecosystem does not yet exist, even as companies rush to ink agreements with their chosen evaluators.
Read the full abstract

All authors Uma Kalkar, Rumman Chowdhury, Reva Schwartz, Gabriella Waters

Dynamical Systems

Linear Response for Self-Consistent Transfer Operators on Graphons

We develop the response theory of the equilibrium state of a large heterogeneous network of weakly coupled uniformly expanding circle maps. The thermodynamic limit of such a network is governed by a Self-Consistent Transfer Operator (STO) acting on measures fibred over a graphon $W$, as introduced in \cite{MPT2025}. We prove that in the weak coupling regime the unique attracting fixed point $\phi^\star(\alpha)$ of the STO depends on the coupling strength in a Lipschitz, and then differentiable, way in a scale of fibrewise Sobolev spaces, and we derive a linear response formula. Our main result is that at zero coupling the response factorises: its derivative along the fibre $z$ equals $\barW(z)\,\Up$, where $\barW$ is the degree function of the graphon and $\Up=-(\Id-P)^{-1}[P(H_\rho\rho)]'$ is a universal profile depending only on the local dynamics and on the coupling function. To first order the heterogeneity of the statistical response across the network is therefore dictated by the degree profile alone, and the finer topology is invisible; the susceptibility of a node is proportional to its degree, so hubs respond most. The topology enters at higher order through a tree expansion: the perturbative coefficients of the equilibrium, defined unconditionally by an explicit recursion, are combinations — with universal coefficient functions — of the rooted homomorphism densities of trees with $m$ edges. We prove that the first two coefficients are the genuine asymptotic ones, the second producing the mean degree of the neighbours of $z$. We also obtain a response formula for perturbations of the graphon itself, Green–Kubo expressions for observables, and the prediction that in a finite network the linear susceptibility of a node is proportional to its normalised degree.
Read the full abstract

High Energy Physics (Phenomenology)

Pion and kaon electromagnetic form factors in a flavor-dressed contact interaction: the role of Nf-screening on dynamical chiral symmetry breaking

We compute the electromagnetic form factors of the pion and the kaon in a symmetry-preserving contact interaction whose effective coupling depends on the number of active light flavors Nf , so that dynamical chiral symmetry breaking is screened as Nf grows toward the chiral transition. The calculation uses the rainbow-ladder truncation of the Schwinger-Dyson and Bethe-Salpeter equations with proper-time regularization, and it includes the dressed quark-photon vertex obtained from its own inhomogeneous Bethe-Salpeter equation in the same truncation. At Nf = 2 the model reproduces the established contact interaction values of the meson masses, decay constants and condensates, and the charge radii rpi = 0.447 fm and rK = 0.417 fm, which undershoot experiment by about one third, as is known for this class of interactions. We therefore restrict every comparison to Q2 less then or equal to 0.6 GeV2, confront it with the NA7 data and with one-loop chiral perturbation theory, and focus on the response to Nf . Both radii grow monotonically with Nf , and the growth survives every variation of the screening function and of the infrared confinement scale that we test. The growth is carried by the quark core rather than by the vector-meson dressing: between Nf = 2 and 6 the bare-vertex contribution to <r2pi> rises from 0.097 to 0.226 fm2, while the dressing contribution moves from 0.102 to 0.127 fm2 even though the rho pole drops from 0.93 to 0.70 GeV. The finite strange mass slows the kaon response only moderately.
Read the full abstract

All authors Aftab Ahmad, Muhammad Abid, Niaz Hussain, Mumtaz Khan, Tariq Mahmood, Alfredo Raya

Artificial Intelligence

AutoEMatic: Runtime-Adaptive Mixed-Precision LLM Inference via Automatic Exponent-Mantissa Refinement

Existing LLM post-training quantization techniques determine precision before inference begins, yet the precision a model requires varies dynamically across prompts, layers, and inference phases. We present AutoEMatic, a full-stack co-design that adapts the precision of weights, activations, and the KV cache at runtime. A lightweight policy selects each layer's precision at every prefill and decode step from two runtime signals: the entropy of the next-token distribution and the magnitude of each layer's update to its hidden states. This runtime adaptation is made practical by AutoFP, a multi-scale floating-point format that derives 2-, 4-, and 6-bit precisions from a single NVFP4 base format. We realize AutoEMatic on two targets: GPU kernels that execute AutoFP on NVFP4 Tensor Cores, and an accelerator whose vector MAC datapath executes all AutoFP formats natively. Across reasoning and agentic benchmarks, AutoEMatic improves accuracy by up to 21.7% over NVFP4 and matches group-scaled FP6 accuracy on most benchmarks, using 4.9 bits per weight versus 4.5 for NVFP4. On GB300 GPUs, AutoEMatic retains approximately 90-98% of NVFP4 decode throughput across the evaluated configurations. Furthermore, the AutoEMatic accelerator reduces latency by 1.6-2.3x and energy by 7-23% compared to an accelerator with NVFP4-only datapath at iso-accuracy.
Read the full abstract

All authors Akshat Ramachandran, Reena Elangovan, Rangharajan Venkatesan, Charbel Sakr, Tijmen Blankevoort, Brucek Khailany, Tushar Krishna

Nuclear Experiment

Probing deformed nuclei: Experimental insights into excited states of $^{152,153,154}$Gd isotopes through fusion-evaporation reactions

The rare-earth region has been the focus of various studies aiming at the understanding of nuclear structure and providing information on the details of the reaction mechanism. The Gd isotopes belong to this group of nuclei and despite the available spectroscopic information, several open questions about their structure still exist, such as the inter-band transitions related to shape evolution or branching-ratios in deformed states. In addition, production cross sections of different reactions to Gd isotopes are largely unknown. In this work, we report on an experimental attempt to populate the excited states in the isotopes $^{152,153,154}$Gd by employing the heavy-ion fusion reaction $^{18}$O+$^{138}$Ba $\rightarrow$ $^{156-x}$Gd + $xn$ in the 58-64 MeV energy range (center-of-mass). The experiment was conducted at the 9 MV FV Pelletron Tandem at the Horia Hulubei National Institute for Physics and Nuclear Engineering, employing the ROSPHERE array. Several branching-ratios for energy levels in $^{152,153}$Gd have been measured, offering new and updated values. Furthermore, relative cross sections regarding the fusion-evaporation reactions $^{138}$Ba($^{18}$O, 4n)$^{152}$Gd, $^{138}$Ba($^{18}$O, 3n)$^{153}$Gd, and $^{138}$Ba($^{18}$O, 2n)$^{154}$Gd have been measured and compared with theoretical calculations with PACE4.
Read the full abstract

All authors S. Pelonis, T. J. Mertzimekis, A. Chalil, P. Vasileiou, A. Zyriliou, G. Zagoraios, D. Papaioannou, F. C. L. Crespi, A. Bracco, N. Florea, N. Marginean, L. Stan, A. Turturica

Representation Theory

Restrictions to a principal $A_1$-subgroup

We consider simple simply-connected algebraic groups $G$ over the complex numbers, and a principal $A_1$-subgroup $A \leq G$. We show that for most types of $G$ all its irreducible representations $V(\lambda)$ ($\lambda$ a dominant weight) are determined by their restrictions to $A$ up to graph automorphisms. The exceptions are groups $G$ of type $A_l$ with $l \geq 5$ and $G$ of type $D_l$ with $l = 4^m$ and $m \geq 2$.
Read the full abstract

Audio and Speech Processing

M5-TTS: A Mini-Scale Multilingual and Multi-Dialect Text-to-Speech Language Model with Multi-Token Prediction

We present M5-TTS, a compact text-to-speech system supporting multilingual and Chinese multidialect synthesis, zero-shot voice cloning, speaker-specific multilingual generation, and CPU streaming. Its 50M-parameter language model, excluding the speech codec, combines a temporal backbone with a local multi-token prediction module over 16 codebooks. GRPO post-training combines global speaker-similarity and pause feedback with a token-level CER process advantage. Character-level error comparisons provide local feedback, with negative credit propagated backward within a clause and projected onto audio-token positions. M5-TTS achieves 1.37% CER on Seed-ZH and 2.91% WER on Seed-EN, corresponding to relative error reductions of 62.2% and 35.3% over MOSS-TTS-Nano with half the language-model parameters. It also achieves lower content error rates in five of six evaluated MiniMax languages, while difficult Chinese inputs remain a limitation. On four Intel Xeon Gold 6530 cores, it achieves a real-time factor of 0.653 and a warm first-chunk latency of 225ms. A 10-listener evaluation gives M5-TTS higher speaker-similarity and intelligibility ratings than MOSS-TTS-Nano, with a lower naturalness rating. We will release training, post-training, and inference code, together with voice-cloning and speaker-specific checkpoints, to support reproducible research on compact speech generation for on-device voice interaction.
Read the full abstract

All authors Wenhao Li, Jingbin Hu, Shengming Ling, Xiaming Ren, Haoyu Zhang, Jinrui Liang, Hanke Xie, Qirui Zhan, Guobin Ma, Huakang Chen, Chunbo Hao, Jianguo Zheng, Wenkang Hou, Dake Guo, Yuepeng Jiang, Linhan Ma, Yunxiang Chen, Houdun Liu, Su Feng, Chuang Ding, Pengcheng Zhu, Liumeng Xue, Lei Xie

Data Structures and Algorithms

Dynamic Edge Orientation on Low-Arboricity Graphs: Breaking the $\log n$ Barrier

We study the dynamic edge orientation problem introduced in a seminal 1999 paper by Brodal and Fagerberg: given a fully dynamic graph $G$ with arboricity bounded by $\alpha$, maintain an orientation on $G$'s edges so every vertex has out-degree upper bounded by a function of $\alpha$. They provided an algorithm with $O(\alpha + \log n)$ amortized update time that maintains an orientation with maximum out-degree $O(\alpha)$. In the decades since, this algorithm has found widespread applications in dynamic graph algorithms. Nevertheless, it has remained a stubborn open question whether, when $\alpha = o(\log n)$, the $O(\log n)$ upper bound can be improved. We answer this 27-year-old question in the affirmative, showing that Brodal and Fagerberg's algorithm for maintaining an out-degree $O(\alpha)$ orientation runs in $O(\alpha \cdot \log \log n)$ amortized update time, which is an exponentially-improved dependence on $n$. This is the first improvement on Brodal and Fagerberg's upper bound even for the $\alpha=1$ case of dynamic forests. Our crucial technical innovation is to construct a hierarchy of dynamic graphs with a property we call near isolation. We believe that this construction may be useful for related problems.
Read the full abstract

Computer Vision

Improving the Diffusability of Motion Tokenizer

Latent diffusion models paired with a variational auto-encoder (VAE) tokenizer have shown promising performance for efficient text-driven human motion generation. However, motion latent diffusion suffers from a generation-reconstruction trade-off: simply increasing tokenizer capacity enhances reconstruction but fails to proportionally improve text-conditioned generation quality. We assume this trade-off stems from a spectral mismatch inside the VAE tokenizer: the encoder yields a latent space biased toward high frequencies, and the decoder introduces temporal artifacts. In this paper, we propose the Diffusable Motion Tokenizer (DiMoT) to improve the diffusability of the tokenizer, rather than scaling up the diffusion backbone. To improve text-conditioned optimization, Critic Spectral Adaptation (CSA) leverages a pretrained diffusion model as the critic to suppress high-frequency latent energy. To preserve motion dynamics, the Alpha-Flow Motion Decoder (AMD) performs generative decoding via average-velocity prediction, enabling efficient one-step decoding. Extensive experiments demonstrate that DiMoT achieves state-of-the-art performance across various challenging text-to-motion benchmarks, yielding superior motion fidelity and text-motion alignment. Project page: https://githubhgh.github.io/dimot/.
Read the full abstract

All authors Guanhe Huang, Songqiao Han, Tangzheng Lian, Oya Celiktutan

Differential Geometry

On the index of two-dimensional elliptic end-periodic operators

This paper investigates the index problem for elliptic end-periodic differential operators on a two-dimensional manifold with a cylindrical end. We define the topological index as the sum of the contribution of the interior symbol, expressed by an Atiyah–Singer type integral, and the contribution of the limit operator, expressed in terms of the Melrose type eta-invariant and a correction term that depends on the lower order terms of the limit operator. We prove the homotopy invariance of the topological index, which allows us to reduce the original problem to the study of operators of a simpler form by employing $K$-theoretic methods.
Read the full abstract

Information Retrieval

Edge-Deployable LLMs for Classification and Value-of-Information Ranking of Disaster-Related Tweets

Humanitarian-aid and disaster relief decisions often rely on noisy, incomplete, and uncertain information from social media. Although substantial research has focused on automating the classification of such data, limited work has examined whether individual posts contain sufficient information to support decisions to act. In this study, we use large language models (LLMs) to (1) classify social media content and (2) assess its Value-of-Information (VoI) in terms of operational actionability. We evaluate seven LLMs for the classification of 77,484 tweets across 10 humanitarian-aid categories and identify the Qwen3 (8B parameters) model as the top-performing classifier, achieving the highest F1 scores in all but one category. In a second task, the classified categories serve as contextual inputs for assessing VoI with respect to responder actionability across six well established metrics for each tweet. We observed that the Gemma 3 (4B parameters) model outperformed the other models on this task, achieving the highest recall across the majority of categories. This work demonstrates that relatively small, edge-deployable LLMs can prioritize actionable information in resource-constrained environments. By integrating VoI assessment, our framework advances the automated elicitation of high-utility knowledge from unstructured social media streams for disaster-response decision-making and highlights a pathway toward tactical edge applications using LLMs that fit on a single 8 GB GPU.
Read the full abstract

All authors Christopher D. Molek, K. Brent Venable, Niranjan Suri, Stephen Russell

Cryptography and Security

BACGUARD: Fact-Grounded Runtime Mitigation of Broken Access Control

Broken Access Control (BAC) has been the foremost category in the OWASP Top Ten since 2021, primarily due to the challenges in enforcing authorization at the application layer. Traditional mechanisms depend on logic embedded in the application, referencing relational database information such as permissions, roles, or ownership. However, missing or incorrect checks can go undetected by external systems, leaving applications vulnerable. This paper presents BACGUARD, a reverse proxy designed to enforce authorization for unmodified applications. BACGUARD reconstructs authorization relations by leveraging authenticated user data from session stores and object ownership from relational schemas. It evaluates whether users are permitted to access resources and blocks unauthorized requests in real time. Unlike existing approaches, BACGUARD bases its authorization on database joins, making the system resilient to forgery. For most requests, decisions are made deterministically. In cases where access control violations are only apparent in response data, such as cross-user data leakage, a language model is used as a fallback. Evaluations on six applications, including those with known vulnerabilities, demonstrate that BACGUARD's hybrid deterministic and language model-based approach blocks all tested attacks with a false-positive rate below 0.7%. Competing solutions like Web Application Firewalls are less effective and more disruptive. The results highlight BACGUARD's primary contribution: the external reconstruction of authorization facts, enabling robust enforcement even when the application itself is flawed.
Read the full abstract