Peer Revenue Publish or profit

Papers

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

All of arXiv

Friday 2 October

Theoretical Economics

Uniqueness of Recursive Utility with Unbounded Consumption and Non-Expected Utility

Luigi Montrucchio and Lorenzo Stanca

We derive sufficient conditions for the existence, non-existence, uniqueness and global attractivity of stochastic recursive utilities. Our approach uses abstract order-theoretic fixed-point arguments, allowing us to accommodate a large class of Markov consumption-growth models with unbounded per-period consumption and a broad range of recursive preferences, including non-expected utility specifications. We illustrate the usefulness of the framework through macro-finance applications to widely used consumption-growth models under several preference specifications, including Epstein–Zin preferences, multiple priors, smooth ambiguity, and Hansen–Sargent robust-control preferences. In each application, our general conditions reduce to a simple, economically interpretable parameter test that shows how fundamental parameters influence existence and uniqueness. While uniqueness of recursive utility can be difficult to guarantee in unbounded environments, in a Markov setting with non-unit elasticity we establish uniqueness within the class of recursive utilities having a so-called separability property.
Read the full abstract

Computational Engineering

Data-Free Weak-Form Staggered Neural Operators for Magneto-Mechanical Coupling in Finite-Strain Elastomers

Alireza Yazdandousthamedani, Ahmad Moeineddin, Reza Najian Asl et al. (+2 more)

Magneto-active elastomers, as a class of smart materials, exhibit strongly coupled magnetic and mechanical behavior at finite strains. Considering variations in microstructure, material properties, and geometry can lead to computationally expensive analyses. Building on the finite operator learning (FOL) framework, this study develops a data-free, physics-informed operator-learning framework for families of coupled finite-strain magneto-mechanical boundary-value problems. The governing magnetostatic and mechanical equations are enforced through finite-element weak-form residuals, allowing the neural operators to be trained without labeled finite-element solution data. The main contribution of this study is the development of a weak-form staggered neural operator (WSNO) framework for strongly coupled magneto-mechanical saddle-point problems. Separate magnetic and mechanical neural operators are trained alternately using a staggered optimization strategy, while their physical coupling is retained through the constitutive relations and residual evaluations. The resulting framework learns mappings from parameterized material and geometric descriptions to the corresponding coupled magnetic and mechanical solution fields. The proposed framework is investigated across several settings, including heterogeneous random-inclusion microstructures, varying magnetic phase contrast, area-fraction-dependent geometries, strongly out-of-distribution material morphologies, and three-dimensional geometry-parametric problems. In addition, the learned operator is combined with a neural-initialized Newton strategy, in which the nonlinear finite-element solver is initialized using the neural prediction. The results demonstrate that the proposed operator-learning framework can accurately capture coupled magneto-mechanical responses across a broad range of parametric problem settings.
Read the full abstract

All authors Alireza Yazdandousthamedani, Ahmad Moeineddin, Reza Najian Asl, Shahed Rezaei, Michael Kaliske

Neural and Evolutionary Computing

Aggregate accuracy conceals concentrated temporal vulnerability in a spiking speech classifier

İsmail Can Dikmen

Aggregate accuracy cannot reveal which utterances are locally vulnerable or how internal activity changes when labels remain stable. We retain every prediction for 725,070 adjacent-bin, one-count changes around 100 validation utterances of a frozen SpikeSCR-based classifier. The canonical native-horizon GPU singleton path reaches 86.0836% validation accuracy. Equal-source expected accuracy under a uniformly chosen neighbor rises from 84.00% to 84.54%, although 13 of 84 initially correct sources admit adverse neighbors. Five sources carry 93.21% of adverse moves. Margin-guided and surrogate-gradient searches miss sparse cases at fixed query budgets; a post hoc gradient prefix finds all 13 at 73.45% of the census of initially correct sources. Source-matched traces and clean-state interventions distinguish internal change from harmful direction and recoverability. Two count-readout replicas have similar validation accuracies but a 5.21-fold class-change gap concentrated in four sources. Controlled query/key source isolation eliminates 531 batch-order label changes and restores bitwise order invariance across all 9,981 validation score vectors. Isolated batch predictions reproduce padding-matched singleton labels on 9,980 of 9,981 inputs. Paired CPU/GPU replay of all 25,820 class-changing neighbors gives 99.8993% label agreement and preserves all 13 vulnerable sources and their fixed witnesses. Complete source-conditioned maps distinguish vulnerability incidence, concentration, internal change, and execution dependence that aggregate accuracy leaves unresolved.
Read the full abstract

Computer Vision

Does Physics Live in the Activations? Localizing Physical Quantities in Video Diffusion Models

Jonas Kneifl, Jakub Skalski, Bartłomiej Twardowski et al. (+1 more)

Video generation models produce strikingly realistic sequences and are increasingly proposed as world models, yet recent benchmarks reveal pronounced deficits in their physical reasoning. This raises the question of whether these models internalize physical principles or merely reproduce familiar motion patterns. We address this by probing internal representations of video Diffusion Transformers (DiTs) for simulator-derived ground-truth physical quantities spanning kinematic motion and rigid-body dynamics under gravity and contact. We find that these quantities are linearly decodable with high accuracy early in the denoising process, substantially outperforming a baseline decoded directly from the model's own noised latents, indicating that the relevant physical information is actively constructed during denoising rather than already present in the input. Additionally, we show that activations at on-object tokens carry the relevant physical information and that quantities defined over multiple frames are readable from single latent frames. Hence, information is sharply localized within the token sequence and is computed globally but stored locally. The probes further show partial extrapolation, transferring to scene variations and object configurations outside their training regime, so what they read is not simply a correlate of the scenes they were fit on. When fitted directly in the full-resolution activation space, the probing directions can serve as steering vectors to change the model's output.
Read the full abstract

All authors Jonas Kneifl, Jakub Skalski, Bartłomiej Twardowski, Kamil Deja

Cryptography and Security

EvoRiskBench: An Evolving Benchmark for Runtime Security Risks in Workspace Agents

Shiyi Kuang, Xuemei Luo, Kun Liu et al. (+7 more)

Workspace agents combine large language models with execution harnesses to perform stateful, multi-step tasks that access or modify external resources. Existing benchmarks leave gaps in executable coverage of their runtime security risks, while evolving model capabilities, harnesses, tools, and threats motivate benchmark evolution. We introduce EvoRiskBench, an evolving benchmark organized around the EP-Path-EF framework, which links an initial risk entry point to a one-hop technical effect through an agent-mediated risk path. The framework defines nine entry-point categories and five effect categories; a 20-participant study supports their interpretability and classification consistency on representative cases. Guided by this framework, an automated end-to-end workflow constructs and executes risk cases in isolated environments and independently verifies outcomes using runtime traces and environment states. The benchmark provides a reproducible dataset of 450 adversarial tasks across six scenarios. We evaluate nine model-harness configurations spanning three models (GPT-5.6 Sol, DeepSeek-V4-Pro-0813, and Claude Opus 5) and three harnesses (Claude Code, Codex, and OpenClaw). Our results reveal substantial vulnerabilities across systems. The most vulnerable configuration, Codex with DeepSeek-V4-Pro-0813, reaches a 68.44% attack success rate (ASR), indicating that configuration of workspace agent is insufficient to ensure secure autonomous execution. ASR varies more across models than harnesses, and harness differences depend on the model. The benchmark cases and evaluation platform will be released after completion of artifact safety and reproducibility checks.
Read the full abstract

All authors Shiyi Kuang, Xuemei Luo, Kun Liu, Junhai Li, Rui Tian, Feng Shi, Bo Shen, Nianyu Li, Dehui Li, Ping Chen

High Energy Physics (Phenomenology)

Some Mathematical Methods Relevant to Analytical Calculations of Anomalous Magnetic Moments of Leptons

L.P. Kaptari, V.I. Lashkevich and O.P. Solovtsova

The mathematical methods and approaches used in our analytical calculations of electromagnetic corrections to the leptonic anomaly $a_L$, which is caused by a special class of Feynman diagrams with insertions of the vacuum polarization operator, are presented and examined in detail. The developed methods are essentially based on the Mellin–Barnes transform for the propagators of massive particles with subsequent application of the PSLQ algorithm to solve the emerging functional equations. Functional equations are also used in the procedure of analytic continuation of a particular type of hypergeometric functions to the entire complex plane of their parameters. It is shown that these approaches make it possible to obtain explicit expressions for radiative corrections of rather high orders.
Read the full abstract

Group Theory

Arithmetic properties of solubilizers

Luca Sabatini

Let $G$ be a finite group and $x \in G$. In 2013, Hai-Reuven proved that the cardinality of the centralizer of $x$ divides the cardinality of the solubilizer of $x$. Settling a conjecture of Mousavi, Poozesh and Zamani, we show that the cardinality of the normalizer of $\langle x \rangle$ is also a divisor. We actually give direct short proofs of more general results.
Read the full abstract

Quantum Physics

Self-organized Layered Structures of Nitrogen-Vacancy Centers with Preferential Orientation in Heteroepitaxial Diamond Films

Vadim Lebedev, Alejandro Martínez-Méndez, Jesús Moreno-Meseguer et al. (+6 more)

Here, we report the fabrication and characterization of single crystal diamond films on a two-inch wafer-scale containing three-dimensional layered structures of negatively charged nitrogen-vacancy color centers (NV$^{-}$). The (100)-oriented layered diamond films were grown by chemical vapor deposition (CVD) to create spatially separated regions of enhanced emission associated with NV$^{-}$ centers. Such localized emission enhancement as well as 2D and 3D arrays of spins could enable close-packed architectures for emerging quantum devices, such as data processors, memory cells, and magnetic sensors. A cyclic sequence of CVD growth and oxygen-plasma etching is employed, resulting in the in-situ formation of layered inverted pyramidal structures with flat $\{$111$\}$ sidewalls. These pits are subsequently overgrown by nitrogen-doped diamond to achieve NV$^{-}$-rich facets with a uniform N-V bond arrangement and improved nitrogen incorporation efficiency.
Read the full abstract

All authors Vadim Lebedev, Alejandro Martínez-Méndez, Jesús Moreno-Meseguer, Jan Engels, Jürgen Weippert, Jan Kustermann, Felix Hoffmann, Javier Prior, Jan Jeske

Neural and Evolutionary Computing

Learning While Inferring: Local and Parallel Learning for Edge SNNs across Sensing Modalities

Yanxun Zhang, Yifei Wang, Changze Lv et al. (+6 more)

Edge intelligence requires models to sense continuously in real time and to keep adapting on-device, all under tight compute, energy, and memory budgets. Although spiking neural networks (SNNs) enable efficient event-driven inference, standard surrogate-gradient backpropagation (BP) serializes updates and blocks ongoing inference. We investigate Bidirectional Spike-Based Distillation (BSD) as an on-device learning principle that lets edge SNNs learn while inferring. BSD couples a stimulus-driven forward network with an independent target-driven reverse network and aligns their intermediate representations through local objectives. Because the two pathways have disjoint computation graphs until alignment, forward inference, reverse inference, and stage-wise updates can run concurrently. On 25 benchmarks spanning the five sensing modalities of SOUL, BSD stays within 3.8 percentage points of matched BP baselines on average, while only its forward branch is needed at deployment. The learned representations also transfer well to few-shot class-incremental learning without replaying past data. By removing the dense floating-point backward chain, BSD reduces the projected training latency to $0.72\times$ and the estimated training energy to $0.36\times$ that of BP.
Read the full abstract

All authors Yanxun Zhang, Yifei Wang, Changze Lv, Jingwen Xu, Yiyang Lu, Xiaohua Wang, Di Yu, Xin Du, Xiaoqing Zheng

High Energy Astrophysical Phenomena

Twins in the Shadow: Phase Transitions in Proto-Neutron Stars

Jannik Milthaler, Jan-Erik Christian, Pia Jakobus et al. (+1 more)

We study the impact of a first-order phase transition from hadronic to deconfined quark matter on the evolution of proto-neutron stars. To maximize the applicability of our results, we construct an equation of state that is highly parameterizable, where we use a relativistic mean field approach to model hadronic matter and a constant speed of sound approach to model quark matter. We approximate the cooling behavior of proto-neutron stars as a sequence of decreasing entropy at constant baryonic mass. Some parameterizations lead to so-called "twin-stars", where two neutron stars have the same (gravitational) mass but different radii. Detecting them is regarded as a smoking gun signal for phase transition. We find, however, that proto-neutron star evolution suggests that gravitational twin stars are only possible if the hybrid star does not have, in addition, a baryonic twin. This rules out most twin star pairs, especially if the phase transition occurs at large densities. In a mass-radius diagram the unpopulated hybrid branch appears to be "in the shadow" of the hadronic branch.
Read the full abstract

All authors Jannik Milthaler, Jan-Erik Christian, Pia Jakobus, Stephan Rosswog

Databases

TSGuard: A Real-Time Framework for Detecting and Imputing Missing Data in Streaming Time Series

Imane Hocine, Asma Abboura, Soror Sahri et al. (+3 more)

Streaming sensor applications routinely suffer from delayed or missing observations caused by faults, communication losses, or environmental interference. Although recent imputation methods exploit temporal and spatial dependencies effectively, most either assume offline access to future observations or prioritize throughput without enforcing domain plausibility. We present TSGuard, a real-time demonstration system for monitoring, validating, and imputing missing values in streaming time series. TSGuard combines a lightweight graph-aware temporal imputation model with constraint-aware validation, fallback estimation, and operator-facing explanations. Rather than treating imputation as an isolated prediction task, TSGuard integrates it into a broader data-quality loop: detect problematic observations, impute missing values, validate estimated against physical and spatial constraints, and either retain the original value as a plausible anomaly or replace it when it violates domain constraints. Using environmental sensing as a motivating setting, the demo enables users to inspect delayed sensors, compare imputers, define constraints, and validate flagged values in real time. The combination of lightweight online spatiotemporal imputation, domain-aware validation, and explicit retain-or-replace decisions is our central contribution, while interactive explanations make these decisions inspectable and actionable. for operators.
Read the full abstract

All authors Imane Hocine, Asma Abboura, Soror Sahri, Abhijith Senthilkumar, Yacine Hakimi, Grégoire Danoy

Theoretical Economics

Signaling via Money

Olivier Bos and Martin Pollrich

We study mechanisms in which agents' preferences depend both on an object allocation and on a third party's beliefs about private information. A mechanism determines an allocation, transfers, and sends signals to the aftermarket. We compare two informational environments: (i) private transfers, where the aftermarket observes allocations and signals, and (ii) public transfers, where transfers are publicly observable. Our main result shows that any equilibrium outcome implementable with hidden transfers can be approximated arbitrarily well with public transfers. Thus, transfer observability generates at most a vanishing informational distortion and helps rationalize the common assumption that transfers remain hidden from aftermarkets.
Read the full abstract

Strongly Correlated Electrons

Correlated Metals, Metamagnetism, and Orbital Nematic Order in moiré Materials with Neural Quantum States

Wei Zhang, Ataç İmamoğlu, Tao Shi et al. (+2 more)

In this work, we investigate a continuum model of two-dimensional electrons interacting via long-range Coulomb forces and subject to a triangular periodic potential at filling factor $\nu=2$. Using neural quantum states based on message-passing neural networks optimized with variational Monte Carlo, we determine the ground-state phase diagram as a function of periodic potential strength and Zeeman field. Although the noninteracting system is expected to form a conventional band insulator, we show that strong correlations can qualitatively reshape the insulating state and, in part of the phase diagram, suppress the band insulator in favor of a correlated metallic phase, with a first-order transition separating the two regimes. Moreover, the metal–insulator transition exhibits a strong dependence on spin polarization: the unpolarized system becomes insulating at substantially weaker periodic potentials through the formation of an $s$-wave molecular insulator, where opposite-spin electrons bind near each potential minimum, whereas the fully polarized system requires a significantly deeper potential and instead forms a symmetry-broken orbital nematic molecular crystal. This spin-dependent molecular transition gives rise to discontinuous magnetization curves, metamagnetic transitions, and several multicritical points in the phase diagram, together with clear signatures in the density profiles, real-space correlation functions, and structure factors. To further elucidate the role of the long-range character of the Coulomb interaction, we also study short-range repulsive interactions, which can be realized with ultracold-atom platforms. We find that the qualitative structure of the many-body phase diagram remains unchanged, demonstrating the robustness of our results across different interaction models.
Read the full abstract

All authors Wei Zhang, Ataç İmamoğlu, Tao Shi, Eugene A. Demler, Ivan Morera

Numerical Analysis

A Physics-Driven Multiscale Method for High-Frequency Helmholtz Equations

Huangxin Chen, Jianhui Chen, Shubin Fu et al. (+1 more)

We develop and analyze a physics-driven multiscale method for model reduction of the Helmholtz equation at high frequency. The method constructs a reduced approximation space from local wave responses within a mixed finite element formulation with velocity elimination. These responses are computed by solving Helmholtz problems with boundary excitations on oversampled regions, giving snapshots that incorporate the oscillatory behavior and propagation characteristics of the underlying medium. Local generalized eigenvalue problems then select modes from this physically informed space to form a reduced global system with sparse interelement coupling. To establish stability and convergence of this reduction, we relate the local spectral approximation to approximation of the global adjoint problem. Starting from an inf–sup estimate for the fine-grid formulation, this argument yields a multiscale inf–sup condition governed by fine-grid resolution, oversampling, and the first omitted local eigenvalues. The resulting stability establishes well-posedness and quasi-optimality of the multiscale approximation in a discrete energy norm. Together with local spectral estimates, it gives an a priori error bound that explicitly quantifies the roles of the coarse and fine mesh sizes, oversampling depth, spectral truncation, and wavenumber. This analysis provides a quantitative link between local space reduction and the stability and accuracy of the global approximation. Numerical experiments in two and three dimensions demonstrate accurate wavefield approximation with substantial reductions in global dimension and solution cost for homogeneous and strongly heterogeneous media, including perfectly matched layer truncations.
Read the full abstract

All authors Huangxin Chen, Jianhui Chen, Shubin Fu, Zhiyu Tan

High Energy Physics (Experiment)

First observation of electron antineutrinos from nuclear reactors at Super-Kamiokande

K. Abe, Y. Asaoka, M. Harada et al. (+246 more)

In 2020, Super-Kamiokande entered the SK-Gd phase, improving its sensitivity to low-energy electron antineutrinos through the detection of neutrons captured on gadolinium dissolved in the detector water. An analysis of reactor electron antineutrino data has been carried out using data from SK-Gd, with a total exposure of 411.52 live days. The observed prompt energy spectra and monthly event rates are consistent with the oscillated reactor antineutrino prediction after accounting for accidental, geoneutrino, and spallation backgrounds. The no-reactor hypothesis is disfavored with a reactor-antineutrino signal significance of approximately $6\sigma$. The oscillation parameters are measured to be $\sin^2\theta_{12} = 0.500\pm0.155$ and $\Delta m^2_{21} = (9.08_{-0.54}^{+0.55}) \times 10^{-5} \mathrm{eV}^2$. A comparison with Super-Kamiokande solar neutrino measurements shows a statistical compatibility between the two datasets at the $1.9\sigma$ level. A combined solar and reactor fit in Super-Kamiokande yields best-fit values of $\sin^2\theta_{12} = 0.332_{-0.026}^{+0.028}$ and $\Delta m^2_{21} = (8.94_{-0.47}^{+0.46}) \times 10^{-5} \mathrm{eV}^2$.
Read the full abstract

All authors K. Abe, Y. Asaoka, M. Harada, Y. Hayato, K. Hiraide, T. H. Hung, K. Ieki, M. Ikeda, J. Kameda, Y. Kataoka, S. Mine, M. Miura, S. Moriyama, K. Nakagiri, M. Nakahata, S. Nakayama, Y. Noguchi, G. Pronost, K. Sato, H. Sekiya, R. Shinoda, M. Shiozawa, Y. Suzuki, A. Takeda, Y. Takemoto, H. Tanaka, S. Chen, Y. Itow, T. Kajita, R. Nishijima, K. Okumura, T. Tashiro, T. Tomiya, X. Wang, F. J. de Garay Arcones, P. Fernandez, L. Labarga, D. Samudio, C. Yanagisawa, B. Jargowsky, E. Kearns, J. Mirabito, L. Wan, T. Wester, B. W. Pointon, J. Bian, B. Cortez, N. J. Griskevich, Y. Jiang, M. B. Smy, H. W. Sobel, V. Takhistov, J. Hill, B. D. Xu, D. Jung, D. H. Moon, R. G. Park, B. S. Yang, K. Scholberg, C. W. Walter, O. Drapier, A. Ershova, M. Ferey, Z. Hu, E. Le Blévec, L. Leplumey, Th. A. Mueller, P. Paganini, C. Quach, R. Rogly, T. Nakamura, J. S. Jang, R. P. Litchfield, L. N. Machado, F. J. P. Soler, J. G. Learned, K. Choi, S. Cao, T. V. Ngoc, L. H. V. Anthony, N. W. Prouse, M. Scott, Y. Uchida, V. Berardi, N. F. Calabria, M. G. Catanesi, N. Ospina, E. Radicioni, A. Langella, G. De Rosa, G. T. Burton, G. Collazuol, M. Feltre, M. Mattiazzi, L. Ludovici, M. Gonin, L. Périssé, B. Quilain, M. Fukazawa, S. Horiuchi, A. Kawabata, Y. M. Liu, Y. Maekawa, Y. Nishimura, A. Oka, H. Tanigawa, Z. Utagawa, R. Akutsu, M. Friend, T. Hasegawa, Y. Hino, T. Ishida, T. Kobayashi, T. Matsubara, T. Nakadaira, Y. Oyama, A. Portocarrero Yrey, K. Sakashita, T. Sekiguchi, N. Bhuiyan, F. Di Lodovico, T. Katori, R. Kralik, N. Latham, R. M. Ramsden, V. Siccardi, S. Aoyama, H. Bambara, Y. Inaba, H. Ito, M. Nishigami, T. Sone, A. T. Suzuki, Y. Takeuchi, S. Wada, H. Zhong, J. Feng, L. Feng, N. Fujimoto, S. Han, J. Hikida, M. Kawaue, T. Kikawa, F. Nakanishi, T. Nakaya, R. A. Wendell, S. J. Jenkins, U. Limbu, N. McCauley, M. Fan, M. J. Wilking, Z. Xie, Y. Fukuda, H. Menjo, Y. Yoshioka, J. Lagoda, J. Zalipska, T. Yano, M. Mori, J. Jiang, Y. Asano, K. Hamaguchi, H. Ishino, Y. Koshio, S. Ohshita, T. Tada, T. Ishizuka, G. Barr, D. Barrow, D. Wark, A. Holin, F. Nova, M. Jo, S. Jung, J. Yoo, L. Kneale, T. Peacock, P. Stowell, H. Okazawa, S. M. Lakshmi, S. Hong, E. Kwon, M. W. Lee, J. W. Seo, I. Yu, Y. Ashida, A. K. Ichikawa, S. Kobayashi, K. D. Nakamura, S. Abe, W. Cai, Y. Endo, S. Goto, S. Kodama, Y. Kong, H. Hayasaki, Y. Masaki, Y. Mizuno, T. Muro, K. Nakagiri, Y. Nakajima, M. Sekiyama, N. Taniuchi, T. Yamazumi, M. Yokoyama, P. de Perio, S. Fujita, C. Jesús-Valls, K. Martens, Ll. Marti, A. D. Santos, K. M. Tsui, M. R. Vagins, S. Izumiyama, M. Kuze, R. Matsumoto, C. Ise, M. Ishitsuka, M. Sugo, M. Wako, K. Yamauchi, Y. Nakano, A. Yankelevich, F. Cormier, R. Gaur, M. Hartz, A. Konaka, X. Li, B. R. Smithers, S. Chen, Y. Wu, A. Q. Zhang, B. Zhang, H. Adhikary, M. Girgus, P. Govindaraj, M. Posiadala-Zezula, Y. S. Prabhu, S. B. Boyd, R. Edwards, D. Hadley, M. O'Flaherty, B. Richards, A. Ali, B. Jamieson, C. Bronner, D. Horiguchi, A. Minamino, Y. Sasaki

Computer Vision

CalCErt: Bin-wise Certification of Confidence Calibration in Medical Image Classification

Leo Fillioux, Stergios Christodoulidis, Stergios Christodoulidis et al. (+2 more)

Deep neural networks remain vulnerable to adversarial perturbations, which can distort not only predictions but also confidence scores, undermining uncertainty calibration. While existing certification methods focus on preserving the predicted category, providing guarantees on how calibration behaves under adversarial attacks remains overlooked. In this work, we introduce CalCErt, a simple and efficient post-hoc strategy that certifies bin-wise confidence calibration for any pretrained differentiable classifier. Our approach combines empirical calibration estimates, statistical concentration bounds, and local Lipschitz estimates of the confidence function to derive data-dependent upper bounds on worst-case miscalibration within an $ell_2$-ball of radius R. We evaluate CalCErt across 11 medical image classification tasks and multiple adversarial perturbations, demonstrating substantially higher certified coverage than baseline strategies while maintaining competitive tightness. Our code is available at https://github.com/leofillioux/calcert.
Read the full abstract

All authors Leo Fillioux, Stergios Christodoulidis, Stergios Christodoulidis, Maria Vakalopoulou, Jose Dolz

Computer Vision

Behavior Pack Optimization for Video MLLM Post-Training

Zhaolu Kang, Shiyu Liu, Tailong Luo et al. (+12 more)

Video multimodal large language models (MLLMs) keep climbing video question answering benchmarks, yet shuffling the frames, masking the segment that supports the answer, or occluding the target object barely changes their predictions. The accuracy rests on appearance and language priors, not on the temporal evidence the question asks for. We trace this to the unit of post-training: rewards are computed on a single response to the original clip, so the model is never asked to behave consistently across views. We propose Behavior Pack Optimization (BPO), which replaces the single response with a behavior pack of outputs across counterfactual views chosen by question type, scored jointly. The pack reward asks for stability when the intervention is irrelevant, sensitivity when key evidence is removed, and abstention when no evidence remains. To keep this objective stable at small pack sizes, BPO uses an anchor-relative advantage: the response on the original view serves as a per-prompt reference instead of a group mean over mixed views. On TempCompass, MVBench, and NExT-QA, BPO improves the macro accuracy of Qwen2.5-VL-7B-Instruct by 4.7 pp, the temporal-hard subset by 7.8 pp, and abstention F1 by 20.0 pp over a budget-matched vanilla GRPO baseline from the same SFT checkpoint. The gains transfer to Video-MME, LongVideoBench, and to LLaVA-Video-7B; ablations confirm they follow the view sets, not the rollout count. We hope this pack-level perspective offers a useful starting point for the video MLLM and multimodal post-training community as the field moves toward evidence-grounded video reasoning.
Read the full abstract

All authors Zhaolu Kang, Shiyu Liu, Tailong Luo, Wei Zhang, Yingjie He, Lei Wei, Guansu Wang, Liang He, Siheng Wang, Guangyuan Dong, Jiaqi Su, Shuang Chen, Haoyu Ji, Qishi Zhan, Kaiyue Zhou

Databases

Tracking State Footprints: How Agents Can Transact

Oto Mraz, Rares Şerban, Kyriakos Psarakis et al. (+2 more)

Can AI agents transact? We argue that they must: as multi-agent systems (MASs) increasingly write code, deploy infrastructure, modify databases, and call web services, lost updates or stale reads can be catastrophic. Although MASs increasingly execute plans in parallel, current orchestrators do not track the state that agents read and write. As a result, concurrency anomalies manifest even in simple coding tasks. We frame MAS coordination as a data management problem and propose to describe agents by their state footprint: the state they read and write across their own local context and state, as well as the state of the orchestrator and external systems. We posit that MASs require guarantees similar to those of databases, but providing them raises new challenges and opportunities: unlike database transactions, agents do not read from a fixed schema or an isolated snapshot, and cannot be replayed deterministically upon failure. They can, however, resolve conflicts semantically instead of aborting, enabling new forms of concurrency control and conflict resolution. Towards agents that can transact, we outline a vision for next-generation agent orchestrators and transactional interfaces for external systems to participate in agentic transactions.
Read the full abstract

All authors Oto Mraz, Rares Şerban, Kyriakos Psarakis, Burcu Kulahcioglu Ozkan, Asterios Katsifodimos

Signal Processing

Agentic RF Intelligence: Multi-Timescale 6G Sensing and Reasoning with On-Device Foundation Models

Jaron Fontaine, Jelle De Moerloose, Xander Vanparys et al. (+3 more)

Future 6G systems share a central challenge with Physical AI: combining sensing, reasoning, and action on network-edge infrastructure in rapidly changing physical environments under strict latency, compute, and energy constraints. Addressing this challenge requires multi-timescale intelligence, combining fast perception that tracks wireless phenomena within milliseconds with slower reasoning that directs sensing and adapts network policies and resources over seconds. In this paper, we present a vision for Agentic RF Intelligence based on a dual-loop architecture that decouples fast, locally autonomous wireless sensing (e.g., detecting spectrum dynamics, interference, and signal sources) from slower agentic reasoning and orchestration. We provide an on-device prototype in which the full stack runs on a single NVIDIA Jetson Thor connected to a B200 Mini USRP. The fast loop processes IQ samples using signal-processing tools and wireless physical-layer foundation models (WPFM), while a local Large Language Model (LLM) asynchronously interprets RF events, invokes tools and steers sensing. Our experiments confirm the separation in timescales, as WPFMs operate at millisecond latency, while LLM interactions take seconds to minutes. The fast loop remains operational during agentic reasoning, providing initial evidence for the feasibility of this decoupled architecture. We conclude by outlining extensions toward memory-driven self-improvement, world models, network control, and multi-agent operation.
Read the full abstract

All authors Jaron Fontaine, Jelle De Moerloose, Xander Vanparys, Anton Lambrecht, Eli De Poorter, Adnan Shahid

Algebraic Topology

Equivariant Chow functions and Wonderful Models in type A and B

Roberto Pagaria and Tommaso Scognamiglio

We prove a general inversion formula in the equivariant incidence algebras of geometric lattices that relates the equivariant Chow function with the equivariant super-reduced characteristic function. We apply this result to the case of minimal wonderful models for hyperplane arrangements associated to root systems, with the natural action of the corresponding Weyl group. Applying this formula to type $A$ (i.e. the partition lattices), we recover Getzler's compositional identity in type $A$. In type $B$ (i.e. the signed partition lattice), we recover Henderson's result, expressing the equivariant Chow generating series in terms of the same series in type $A$ and the equivariant characteristic generating series of type $B$. Moreover, we obtain a new compositional identity for the equivariant Chow generating series.
Read the full abstract

Artificial Intelligence

Keeping JEPA World Models Plannable When Little of the Frame Moves

Florian Strohm, Patrick Wagner, Jannik Schwab et al. (+1 more)

Specifying a goal in language rather than as a goal frame is a natural interface for planning with a latent world model, but testing it needs scenes in which language must discriminate between several objects. We build SLIM, a pushing benchmark with several small objects and paired visual and language goals on identical scenes. On SLIM a LeWM world model that solves PushT succeeds on under 1% of trials, although a scripted controller with simulator state solves every tier. Probes locate the failure in the encoder: its latent is nearly action-insensitive, neither pusher nor object positions can be decoded from it, and rollouts are no better than copying the current latent forward. One inverse-dynamics auxiliary loss, applied to encoder latents and to predicted latents through a shared head discarded at test time, restores every probe and raises success from 0.003 to 0.35 (0.16 on the hard pushing tier, where a goal-agnostic policy scores zero), and improves PushT at twice the trained horizon. Controls attribute the repair to the gradient into the encoder, and a response sweep shows that the vanilla model plans once enough of the frame responds to actions. A cheap action-sensitivity probe, computable without environment access, acts as an empirical necessary condition: all configurations below its threshold failed to plan. On the repaired latent, a small language-goal head plans from sentences without retraining the world model: it reaches 0.84 on navigation (visual-goal oracle 1.00), follows the named zone when it is swapped with a decoy, and degrades gracefully to unseen nouns. A single goal sentence rarely completes a push, but given the push as a sequence of stage sentences the head raises success on the medium and hard pushing tiers from 0.04 to 0.25, on par with the goal-frame oracle, also when the switch between stages is read from the latent alone.
Read the full abstract

All authors Florian Strohm, Patrick Wagner, Jannik Schwab, Marco Huber

Computation and Language

Investigating the Role of Reasoning-Language Alignment in Monolingual Retrieval-Augmented Generation

Oliver Hauck, Mario Sanz-Guerrero and Katharina von der Wense

Reasoning traces improve large language models (LLMs), but current models are trained to reason mostly in English. It has been shown that forcing a model to reason in another language degrades accuracy, even when the reasoning language matches the language of the prompt – but only for a setting where the model reasons over a short prompt. Here, we ask whether the same holds for retrieval-augmented generation (RAG), where the model must read and integrate a large amount of retrieved evidence in the target language. To study this, we build a fully monolingual German RAG question-answering testbed over the fictional world of the tabletop role-playing game The Dark Eye, a domain that is richly documented in German but too niche for the model to answer from memory, so that it has to rely on retrieval. Varying the forced reasoning language of an agentic RAG system on this testbed, we find that aligning the reasoning language with the language of the query and the retrieved documents helps. Forced German reasoning outperforms forced French, although the model benchmarks higher in French, so the benefit comes from alignment and not from language proficiency. The advantage grows when the retrieved context is richer and structure-aware. However, forced German only reaches the level of the model's native, unconstrained English reasoning without surpassing it, showing that native multilingual reasoning is needed. We publicly release the testbed and QA benchmark.
Read the full abstract

Machine Learning

Page-EntroKV: Hardware-Aligned, Entropy-Weighted KV-Cache Eviction under Grouped-Query Attention

Inbasekaran S

Serving long-context autoregressive language models is constrained by the key-value (KV) cache. Most dynamic eviction methods score token importance per query head and choose tokens independently. This fits poorly with grouped-query attention (GQA), where several query heads share one physical KV buffer: divergent per-head selections force the serving engine to retain the union of their choices - inflating the cache by up to the group ratio r - while arithmetic mean pooling dilutes the specialized retrieval heads that carry factual recall. We introduce Page-EntroKV, a formal framework for KV-cache eviction operating at the granularity GQA serving actually allocates. Heads within each physical group are pooled by weights derived from sink-isolated collision (Renyi-2) entropy - one inner product per head, computed once at prefill with no calibration - so sink heads cannot masquerade as retrieval heads. Pooled scores are projected onto PagedAttention page frames, and eviction executes at the hardware tuple (layer, group, page). We formalize the union overhead ratio (UOR) and intra-group disagreement, prove an exact identity linking them for two-head groups alongside two-sided bounds at every group ratio, prove strict budget preservation and a finite-context needle-retention bound that arithmetic mean pooling provably violates, and give exact per-layer page accounting. On a pilot architecture (Qwen2.5-1.5B-Instruct, r=6), head-independent replay over 2,240 group measurements yields union overhead up to 4.75x at a 2% budget, while Page-EntroKV holds UOR exactly 1.000; sink isolation removes a 13x sink masquerade; needle recall is 100% versus 0% for mean pooling at a 20% budget; retained cardinality is exact for every page size; and QA and code tasks remain solvable at 20% retention.
Read the full abstract

Number Theory

Pellarin's Identity, Carlitz Period, and Anderson Generating Functions over Arbitrary Curves

Chuangqiang Hu, Stephen S.-T. Yau and Lishan Yu

Let \(A\) be the coefficient ring of a smooth projective curve over \(\mathbb{F}_q\) with a closed point at infinity of arbitrary degree \(\mathrm{N}=\deg(\infty)\ge 1\), and let \(\varphi\) be a rank-one Drinfeld \(A\)-module. In this paper, we prove three main results concerning the module of special functions and its relation to Pellarin's \(\mathcal L(1)\)-series. First, the module \(\mathrm{sf}(\varphi)\) of special functions contains a Tate-algebra unit exactly when it is free of rank one, equivalently when its period lattice is isomorphic to the module of regular differentials. This settles the Gazda–Maurischat conjecture. The proof evaluates Cauchy kernels; on the period lattice all such evaluations agree, and their common characteristic residue recovers the period. Second, for suitable isogeny data to a principal-period target \(\psi\), a strictly normalized shtuka product, together with one characteristic residue, gives an explicit formula for the fundamental period \(\tilde{\pi}\) (the generator of the free lattice \(\Lambda_\psi\)). The same defect equation also yields the residue formula for the Drinfeld logarithm. Third, the first Frobenius twist of the normalized shtuka differential is identified with a pairing built from finite torsion traces, whose coefficients are finite étale traces over the field of definition. As an application, Pellarin's \(\mathcal L(1)\)-series is expressed exactly as the first Frobenius twist of the generating differential at a fundamental period. A notable feature of these results is that they hold for arbitrary $\mathrm{N} \geq 1$, and the source period lattice need not be free.
Read the full abstract

Quantum Physics

Infrared Memory of a Majorana Shutter

Ali Vahedi

We study a finite-width Majorana shutter as a local, time-dependent scattering defect and determine which features of its infrared memory are insensitive to the microscopic spatial profile. For a smooth shutter, the zero-frequency scattering matrix is fixed by the integrated coupling, while finite-width corrections vanish in the long-wavelength limit. A time-dependent switch then produces an infrared Bogoliubov kernel whose singular part is controlled by the change in this scattering matrix. We examine the many-body response through three observables. The overlap of the pre- and post-shutter Gaussian vacua exhibits Anderson orthogonality, with an exponent set by the scattering angle. A lattice calculation of the post-quench quasiparticle number resolves its logarithmic infrared growth, reaching the predicted coefficient in the slow-switch test and reproducing the expected dependence on the scattering-angle change. When the entanglement cut passes through the shutter, the entropy follows the effective-central-charge description of a conformal defect, with parameter-free agreement at the percent level, improving for the wider shutter. When the shutter lies inside the interval, its contribution remains a subleading finite-size correction. A local bilinear correlator approaches the expected Majorana power law and shows no stable logarithmic amplitude. These results show that the same low-frequency scattering data control distinct global and spatial observables, while their infrared coefficients are not interchangeable. The remaining limits are also made explicit: the finite-memory correction to the continuum kernel and the controlled continuum limit of large-interval spatial entropy.
Read the full abstract

Robotics

Safe Streaming Flow Planning by Aligning Sampling Dynamics with Execution Dynamics

Seunghwan Jang, Jeongyong Yang, Siddharth Ancha et al. (+1 more)

Generative planners based on diffusion/flow matching can learn to synthesize long-horizon trajectories from demonstrations. However, real-world deployment requires (i) enforcing safety constraints during execution and (ii) tight online replanning at fast execution rates. Prior safe diffusion/flow planners generate the agent's full trajectory at once, while repeatedly perturbing intermediate states to satisfy safety constraints. This approach is not only computationally intensive, but also introduces distribution shift since the learned sampling dynamics is distinct from the system's execution dynamics. We propose SafeStreamingFlow, a goal-conditioned planner that aligns flow sampling dynamics with execution dynamics by sequentially integrating a learned state vector field with hierarchical state prediction. Importantly, we need to enforce safety constraints only for the executed step via high order control barrier functions. Across navigation, racing, and locomotion benchmarks, SafeStreamingFlow reduces planning latency and improves safety compared to existing methods, while maintaining competitive goal-reaching success.
Read the full abstract

All authors Seunghwan Jang, Jeongyong Yang, Siddharth Ancha, SooJean Han

Combinatorics

A note on partitioning the vertex set of a graph into a dominating set and a locating dominating set

Dipayan Chakraborty, Florent Foucaud, Michael A. Henning et al. (+1 more)

A set $S$ of vertices in a graph $G$ is a dominating set of $G$ if every vertex not in $S$ has a neighbor in $S$, where two vertices are neighbors if they are adjacent. The domination number, $\gamma(G)$, of $G$ is the minimum cardinality among all dominating sets of $G$. Given a set $S$ of vertices of a graph $G$, two vertices are located by $S$ if they have distinct sets of neighbors in $S$. Moreover, if $S$ locates every pair of vertices not in $S$, then it is called a locating set of $G$. A locating dominating set of $G$ is both a dominating and a locating set of $G$. The locating domination number, $\gamma^{\rm LD}(G)$, is the minimum cardinality among all locating dominating sets of $G$. A notable conjecture in the study of locating dominating sets is to show that the locating domination number of an isolate-free and twin-free graph of order $n$ is at most $\frac{1}{2}n$. So far, the best approximation to this upper bound conjecture is known to be $\left \lceil \frac{5}{8}n \right \rceil$. Much in line with the conjecture, an even stronger reformulation proposed in the literature asks if it is possible to partition the vertex set of an isolate-free and twin-free graph into two locating sets. However, such partitions into locating sets may not exist if the graph is also allowed to have twins. Continuing with this line of research, we show that if $G$ is an isolate-free (and not necessarily twin-free) graph, then the vertex set of $G$ can be partitioned into a dominating set and a locating dominating set. As a consequence, we infer that every isolate-free graph $G$ of order $n$ satisfies $\gamma(G) + \gamma^{\rm LD}(G) \le n$, and we show that the last bound is tight. Moreover, our proof of the existence of a partition of the vertex set of an isolate-free graph into a dominating and a locating dominating set also provides a polynomial-time algorithm to construct such a partition.
Read the full abstract

All authors Dipayan Chakraborty, Florent Foucaud, Michael A. Henning, Tero Laihonen

Information Retrieval

Benchmarking Literature Retrieval for a Model Organism: A Dictyostelium Case Study

Yun Wang, Gad Shaulsky, Tomaž Curk et al. (+1 more)

Biological literature retrieval systems are often developed and evaluated using broad biomedical corpora and general-purpose search tasks. However, many curated knowledge bases operate in narrower model-organism domains, where the literature is sparse and terminology is organism-specific. We introduce a retrieval benchmark from dictyBase for Dictyostelium, a model organism in cell and developmental biology. The benchmark consists of curator-generated biological queries linked to PubMed-indexed articles, together with structured gene annotations. Using this benchmark, we study three factors in niche biological retrieval: cross-encoder reranking, gene-aware query expansion, and abstract-only versus full-text retrieval. We report that reranking and gene-aware query expansion improve retrieval selectively: reranking is most useful when the model is well suited to biological evidence matching, whereas curated annotations help clarify compact biological queries by reducing vocabulary mismatch. Full-text chunks substantially improve retrieval when abstracts omit supporting evidence, increasing both candidate recall and top-rank performance, although these cases are harder than queries supported by abstracts. Data and code are publicly available at https://github.com/fulaibaowang/dictycite, and the benchmark dataset is additionally archived on Zenodo.
Read the full abstract

All authors Yun Wang, Gad Shaulsky, Tomaž Curk, Blaž Zupan

Differential Geometry

On the uniqueness of Ricci flows from certain Alexandrov spaces

Laura Flower

We consider the problem of uniqueness of Ricci flows, starting from singular initial data. Even when the initial data is smooth and smoothly obtained, in the complete noncompact case uniqueness is not known in full generality. When starting from nonsmooth initial data, very little is known, even in the compact case. We obtain a uniqueness result for Ricci flows with scaling-invariant curvature bounds, a suitable noncollapsing assumption, and sectional curvature lower bounds, coming out of (possibly singular) Reifenberg spaces. As a consequence of our result, and existence results obtained by Simon–Topping and Lai, there is a canonical way to define a smooth structure on any Reifenberg uniformly volume-noncollapsed PIC2 limit space. Further, we obtain a diffeomorphic version of Perelman's stability theorem for these limit spaces.
Read the full abstract

Artificial Intelligence

Trading Strategy Optimization via Textual Gradient

Chaoqun Yang, Qian Wang, Fengbin Zhu et al. (+4 more)

Quantitative trading strategy design aims to discover trading programs from historical data that remain effective in future markets, which can be viewed as a black-box program optimization problem. LLM-based textual gradients offer a promising approach by providing explicit optimization directions for iterative strategy refinement. However, directly applying textual gradients faces two challenges: (1) optimization is myopic, underutilizing experience from previous evaluations; and (2) aggregate backtest feedback overlooks temporal robustness, potentially favoring strategies that perform well only in specific market periods. To address these challenges, we propose TradeGrad, an experience-guided textual-gradient framework for robust trading strategy optimization. TradeGrad leverages accumulated optimization experience to estimate textual gradients and employs multi-scale revisions for both strategy exploration and refinement. It further introduces the Cross-Period Robust Objective (CPRO), which emphasizes performance in unfavorable historical periods to promote temporal robustness. Experiments on cross-sectional and time-series strategy design in Chinese A-share and U.S. equity markets show that TradeGrad achieves the best in-sample and out-of-sample performance across all four settings. Notably, its Chinese cross-sectional strategy achieves 27.99% annualized return, 12.19% maximum drawdown, and a Sharpe ratio of 1.63, approximately 68% higher than the CSI 300 benchmark. Further analyses validate the proposed components and show consistent improvements in both in-sample and out-of-sample performance throughout optimization. The code is available at https://github.com/transcend-0/TradeGrad.
Read the full abstract

All authors Chaoqun Yang, Qian Wang, Fengbin Zhu, Xinyu Lin, Bingsheng He, Roger Zimmermann, Tat-Seng Chua