Peer Revenue Past citations are no guarantee of future returns

Papers

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

All of arXiv

Monday 14 September

Chaotic Dynamics

Decision-Related Cognitive Signatures from Fast-Slow Dynamics: A Low-Dimensional Observation-Operator Framework

Repeated decisions exhibit temporal structures such as persistence, direction-dependent switching, recurrent alternation, and abrupt transitions. We examine the generative sufficiency of a two-dimensional fast-slow dynamical system. The system combines a cubic fast equation with linear slow feedback and is analyzed through its equilibrium geometry, trace-determinant structure, equilibrium-fold loci, candidate Hopf boundaries, and singular critical manifold. An explicit observation operator projects continuous trajectories to a scalar signal and applies a binary readout, separating latent state-space dynamics from observable behavior. The analysis establishes a unique-equilibrium regime and, for suitable parameters, a three-equilibrium wedge with a central saddle. The outer equilibria are attracting only where their traces are negative. The analysis also identifies the simple-zero condition required for ordinary saddle-nodes, trace-zero positive-determinant spectral boundaries compatible with oscillatory instability, and the attracting and repelling branches of the critical manifold. Prescribed nonautonomous sweeps numerically illustrate direction-dependent switching (T1), transient episodic recurrent switching (T2), and an abrupt localized regime shift (T3). The attracting-equilibrium regime associated with prolonged state retention (T4) is characterized analytically. The resulting correspondence is intended as a test of generative sufficiency at the level of observable temporal organization, rather than as an identification or empirical validation of latent cognitive mechanisms.
Read the full abstract

Metric Geometry

Additive quasi-isometries and cacti

We prove that if a geodesic metric space contains no $c$-fat theta curve for some $c>0$, then it is $(1,K)$-quasi-isometric to a cactus graph, where $K$ depends only on $c$. Using a coarse characterization of cacti in terms of $c$-fat theta curves this implies that every geodesic metric space quasi-isometric to a cactus is $(1,K)$-quasi-isometric to a cactus graph.
Read the full abstract

General Relativity and Quantum Cosmology

Curvture Regularization and Dynamical Vacuum Structure in Pseudo-Complex General Relativity

Whether classical spacetime remains welldefined at arbitrarily high curvature is a central question. In general relativity, singularities signal the breakdown of the classical description, motivating modifications of spacetime's short-distance structure. Pseudocomplex general relativity (pcGR) extends spacetime geometry to pseudo-complex coordinates, naturally introducing two metric sectors, the physical metric g{\mu}{\nu} and an auxiliary field f{\mu}{\nu}. The magnitude of the pseudo imaginary component defines the invariant acceleration scale a0, a direct consequence of the pseudo-complex structure. The simultaneity condition, requiring both idempotent sectors to satisfy the pseudo-complex Einstein equations independently, fixes the auxiliary field algebraically, with no new propagating degrees of freedom, and sets a lower bound on the lapse that regularizes curvature. The regularization is achieved through the combined effect of the lapse-gap condition e{\nu}(r) gt. a0, which ensures the radial metric component remains nondegenerate, together with the regularity conditions at the areal-radius origin, B(0) = 1 and B'(0) = 0, which emerge naturally from the pseudo-complex geometry. These conditions jointly eliminate the Schwarzschildtype curvature divergence, rather than the gap condition acting alone.
Read the full abstract

Machine Learning

Safe Meta-Reinforcement Learning via Information Space Reachability

Meta-reinforcement learning (meta-RL) enables agents to adapt to unseen tasks with limited experience. Despite its promise, the application of meta-RL in real-world tasks is hindered by safety requirements, which have been underexplored in prior work. In this paper, we propose a safe meta-RL framework that explicitly accounts for safety during adaptation. Our key insight is to reason about safety in the information space, which captures both the physical state and the agent's belief over the underlying task. Within this space, we introduce a safety value function that measures the probability of the agent avoiding unsafe regions indefinitely. We show that this function satisfies a self-consistency condition and a Bellman equation, which make it learnable via meta-RL. Based on this formulation, we develop a safe meta-RL algorithm that learns the safety value function and leverages it for safety filtering and constrained policy optimization. Experiments on meta-RL benchmarks demonstrate the effectiveness of the proposed method.
Read the full abstract

Soft Condensed Matter

Dynamically triggered snap-through

Snap-through is a sudden and large mechanical deformation that is usually triggered by driving a control parameter beyond a bifurcation threshold. Here, we study a buckled strip that undergoes a saddle-node bifurcation as one of its clamps is rotated. We show experimentally, theoretically, and numerically that the effective snapping threshold can be markedly lowered by speeding up the rotation. For actuation times that are short but nevertheless longer than the fundamental flexural period, the strip is observed to remain close to the instantaneous quasi-static equilibrium for most of the loading process. This allows us to derive an energetic criterion that accurately estimates a dynamic instability threshold well below the saddle-node bifurcation point. On the other hand, very near the bifurcation point, the evolution of the quasi-static equilibrium undergoes a geometric boost, which entrains the general motion of the beam. As a result, a slower actuation of the clamp than predicted by the energy argument suffices to produce snap-through. A distinct scaling is derived in that limit, based on the aforementioned boost and critical slowing down of the least stable vibration mode of the beam. More broadly, our results show how finite-rate forcing can trigger snap-through before the quasi-static instability threshold is reached
Read the full abstract

All authors Gregory Kozyreff, Lucie Domino, Basile Radisson, Hadrien Bense

Astrophysics of Galaxies

Common Excitation Patterns of Star Formation, Active Galactic Nuclei, and Shocks in Seyfert Galaxies

The growth of galaxies and their central supermassive black holes is closely connected, yet the net effect of active galactic nuclei (AGN) feedback on host-galaxy star formation remains uncertain. AGN may enhance, suppress, or have little measurable impact on star formation, but distinguishing among these outcomes requires separating star-formation and AGN photoionization from shock excitation, which is expected in AGN-driven outflows but has been difficult to isolate. Here we apply a recently developed theoretical three-dimensional diagnostic diagram, designed to separate star formation, AGN, and shock excitation, to VLT/MUSE IFU observations of nine nearby (z < 0.026) Type 2 Seyfert galaxies. We find a common excitation pattern across the sample: star-forming rings or arcs at projected radii of r~0.8-6 kpc, AGN-photoionized bicones extending to kpc scales, central fast-shock-dominated regions that often extend perpendicular to the AGN bicone, and pure-shock- dominated regions surrounding the central fast shocks and appearing locally within the star-forming rings. Deep Chandra X-ray morphology independently supports this decomposition. The circumnuclear star-forming rings are consistent with bar-driven resonances, although positive AGN feedback may also contribute. The central fast shocks are broadly consistent with AGN jet-ISM interactions, while AGN wind-ISM interactions may also play an important role in galaxies with low-power jets. These results establish central shocks as a common feature of Seyfert galaxies and demonstrate the importance of accounting for shock excitation in AGN feedback studies.
Read the full abstract

All authors Peixin Zhu, Lisa J. Kewley, Dominika Ł. Król, Giuseppina Fabbiano, Lars Hernquist, Ralph S. Sutherland, Anna Trindade Falcão, Martin Elvis, Riccardo Middei

Numerical Analysis

Temporal Error Growth of Strang Splitting Method for the Periodic Cubic NLS

We study the temporal error growth of the Strang splitting method for the periodic cubic nonlinear Schrödinger (NLS) equation. The leading error is governed by a forced linearized equation, whose growth depends sharply on dimension and the sign of the nonlinearity. In the 1D defocusing case, we prove a uniform quadratic upper bound $C(1+T^2)\tau^2$ using the global Birkhoff transformation and the resulting degenerate structure of the linearized flow. In the higher-dimensional defocusing case, the exponential growth is constructed using arbitrarily small unstable standing waves. Moreover, to prove the higher-dimensional defocusing upper bound, we use the periodic Strichartz estimates of Killip and Vişan to show that $\int_0^T\|u(t)\|_{L^\infty}^2\,dt\le C(u_0)(1+T)$. This yields an exponential rate independent of the final time, despite possible growth of higher Sobolev norms. The exponential lower bound in the focusing case is obtained from the unstable linearized dynamics around a plane wave, with the Akhmediev breather providing the underlying mechanism. In addition, for 1D focusing case with small initial data, the appliction of the local Birkhoff transformation ensures that the error grows at most quadratically in time. Various numerical experiments confirm the sharp linear, quadratic, and exponential error growth rates.
Read the full abstract

Complex Variables

The Cesaro operator is cyclic on $H^p$

In this paper, we show that the Cesàro operator on the Hardy space $H^p$, $0 < p < \infty$, is cyclic. Our techniques will involve semigroups.
Read the full abstract

All authors Anil Belli, Ugur Gul, William T. Ross, Aristomenis G. Siskakis

Robotics

SlipSense: Multimodal Tactile Learning for Low-Latency and Generalized Slip Detection

Slip detection is fundamental to dexterous manipulation, yet existing systems often lack precise characterization of detection latency and cross-platform generalization. We present SlipSense, a multimodal tactile slip-detection framework built on TacV5, a compact sensor integrating a $32 \times 32$ piezoresistive array operating at 240 Hz and a 3-axis MEMS accelerometer operating at 8 kHz. The piezoresistive array captures spatial pressure distributions, while the accelerometer captures friction-induced vibrations, providing complementary slip cues. The framework performs modality-specific encoding, intra-sensor fusion, and cross-modal attention with causal temporal prediction at 240 Hz. Experiments on a dataset of 1.4 million frames spanning 37 objects demonstrate the complementarity of the two modalities. SlipSense achieves 96.7% Macro F1 with a false-positive rate below 1.6%, detecting 76% of slip events within 23.1 ms. When trained solely on UMI data, SlipSense generalizes zero-shot to a Tesollo dexterous hand, transferring across unseen objects, distinct sensor units, and robotic platforms without retraining.
Read the full abstract

All authors Tong Jian, Aditya Thurvas Senthil Kumar, Xinyi Li, Ziling Chen, Tianyu Dai, Ali Sengul, Matteo Grimaldi, Wenjie Lu, Saleh Nabi, Tao Yu

Probability

Naturality characterizes product measures and relative product measures

We show that the formation of the product measure is the only natural way to assign to each tuple of probability measures on measurable spaces a probability measure on the product space. Here, naturality is meant in the sense of category theory and amounts to the condition that the assignment commutes with pushforward along measurable maps. In the standard Borel setting, we also prove an analogous result for relative products over a fixed base probability space. As a corollary, we conclude that the formation of relative products is a lax symmetric monoidal functor.
Read the full abstract

Mathematical Physics

Sharp decay thresholds for eigenvalues of discrete Schrödinger operators on $\mathbb{Z}^2$

We study eigenvalues of discrete Schrödinger operators $H=\Delta+V$ on $\mathbb{Z}^2$, where $\Delta$ is the uncentered Laplacian, i.e., the un-normalized adjacency operator of $\mathbb{Z}^2$ and $V$ decays at infinity. By Weyl's theorem, the essential spectrum of $H$ is $[-4,4]$. We determine the sharp decay thresholds for existence of eigenvalues in three distinct spectral regimes. While it is natural to expect different behavior at the spectral edge $\lambda=\pm4$, and the bulk, there is a further distinction between the regular energies $0<|\lambda|<4$ and the interior critical energy $\lambda=0$ stemming from the reducibility of the corresponding Fermi surface in the latter case. For every $0<|\lambda|<4$, we construct potentials $V$ satisfying $|V(n)|\leq C|n|^{-1}$ for which $\lambda$ is an eigenvalue of $\Delta+V$, and prove absence of eigenvalues when $|V(n)|\leq C|n|^{-1-\varepsilon}$ for some $\varepsilon>0$. At $\lambda=0$, the critical power changes and we construct potentials $V$ satisfying $|V(n)|\leq C|n| ^{-2}$ for which $0$ is an eigenvalue of $\Delta+V$, as well as prove absence when $|V(n)|\leq C|n|^{-2-\varepsilon}$ for any $\varepsilon>0$. Finally, at each spectral edge, we show that, for every $K\geq3$, an eigenvalue can be created by potentials supported on exactly $K$ sites, whereas a potential supported on at most two sites cannot create an edge eigenvalue. The proofs combine Green-function expansions and moment cancellation, Hilbert-space-valued iterations, discrete Carleman estimates, and a uniform Green-kernel estimate.
Read the full abstract

Differential Geometry

Scalar-mean rigidity theorem and Llarull's theorem for nonspin manifolds

We prove Llarull's scalar curvature rigidity theorem for spheres and scalar-mean rigidity for strictly convex Euclidean domains in all dimensions without the spin assumption.
Read the full abstract

All authors Gaoming Wang, Jinmin Wang, Zhizhang Xie, Bo Zhu

Cryptography and Security

Authorization Architectures for Tool-Using AI Agents

Tool-using artificial intelligence (AI) agents, systems that autonomously invoke application programming interfaces (APIs), databases, browsers, and inter-agent protocols such as the Model Context Protocol (MCP), are becoming production infrastructure. Yet the security model governing when an agent is authorized to act on a human's behalf remains underdeveloped. Trustworthy human-AI systems require that every consequential agent action be traceable to a human principal, bounded by what that human actually delegated, and contestable after the fact; few documented deployments satisfy all three properties reliably and end to end. Existing literature addresses fragments of this problem in isolation, credential management for non-human identities, classical access control models, prompt injection, and audit trails, while giving little attention to the authorization decision point itself, the moment a tool invocation occurs, and mechanisms that make that decision correct, enforceable, and accountable. This review introduces a principal hierarchy spanning human user, operator/deployer, orchestrator agent, sub-agent, and tool endpoint as an organizing framework, and examines five interdependent layers: agent identity and credential lifecycle; delegation and scope propagation across multi-hop chains; runtime enforcement and just-in-time authorization at policy enforcement points (PEPs); prompt injection as an authorization bypass that breaks the principal hierarchy; and auditability, provenance, and non-repudiation. Drawing on a structured narrative review of 89 primary sources screened from approximately 180 candidates published between 2023 and 2026, we propose seven structural requirements, derive a four-layer reference architecture, apply the requirements to three deployable reference configurations, and identify runtime enforcement and aggregation bounds as the principal unresolved gaps.
Read the full abstract

All authors Rakesh Kumar Surapani, Pradeep Kumar Dolabehera Kakitapelli, Arun Morampudi, Praveena Padi

Solar and Stellar Astrophysics

The JWST Proto-PAH Project: Detection of the methyl radical CH$_{3}$ in the highly evolved C-rich object SMP LMC 011

We report the first detection of the neutral methyl radical CH$_3$ in a highly evolved C-rich object, SMP LMC 011, using JWST MIRI/MRS. CH$_3$ is well fitted by an excitation temperature of $T_{\rm ex} \simeq$~190\,K and a column density of $N_{\rm tot}(\mathrm{CH_3}) \simeq 5.6\times10^{17}$ cm$^{-2}$. We also report the non-detection of ethane (C$_2$H$_6$), consistent with CH$_3$ reacting preferentially with unsaturated radicals rather than self-recombination, which may be inefficient in SMP LMC 011. Combined with the exceptionally large benzene column density of SMP LMC 011, this supports a methyl-addition route from benzene towards alkyl-substituted aromatics. We propose that the high CH$_3$ abundance is driven by the erosion of hydrogenated amorphous carbon (HAC) dust grains in the dense warm torus by UV photons from the central star and/or by shocks, which release CH$_3$ directly into the gas phase. These results establish CH$_3$ as a key reactive intermediate in the formation of complex hydrocarbons in C-rich circumstellar environments. Chemical models that incorporate methyl-addition reactions are needed for a better understanding of PAH formation pathways in evolved stars. The potential detection of alkyl-substituted aromatics such as toluene (C$_7$H$_8$) and ethylbenzene (C$_8$H$_{10}$) in C-rich evolved stars would provide direct confirmation of CH$_3$-driven aromatic growth in circumstellar environments.
Read the full abstract

All authors Jialu Li, D. A. García-Hernández, A. Manchado, Debayan Das, J. Cami, Els Peeters, G. C. Sloan, B. Aringer, J. Bernard-Salas, C. Bhatt, Nicholas Clark, Harriet L. Dinerstein, M. A. Gómez-Muñoz, Kathleen E. Kraemer, M. Matsuura, R. Sahai, N. C. Sterling, Kevin Volk, G. M. Wahlgren, A. A. Zijlstra

Human-Computer Interaction

How do people plan digitally: An in-the-wild investigation of task planning through a smartphone app

Daily planning supports goal attainment and productivity, and people increasingly delegate it to digital tools. Plans are postponed, revised, and left unresolved rather than executed as intended. Understanding these changes requires following tasks from creation through later updates and recorded outcomes. We report an in-the-wild study of task planning using time-stamped logs of 24,265 tasks from 957 users of a widely used daily planner app, observed over six weeks alongside self-reported surveys. Following each task across its lifecycle, only 26.4% moved directly from creation to completion and 32.0% were abandoned; all-day and longer tasks were abandoned disproportionately, and 77.6% of timed tasks were marked complete later than intended. Latent profile analysis of 909 users indicated High Engagement (11.1%), Low Engagement (19.7%), and Passive (69.2%) profiles; active days on app distinguished the profiles and were associated with post-survey completion. These findings support evaluating planning tools across the task lifecycle.
Read the full abstract

All authors Srija Halder, Isabella Zimmermann, Linnea Körte, Leo Mehlig, Paul Schmiedmayer, David Grüning

Machine Learning

Discrete Beckmann Transport Models for One-Step Language Modeling and Reasoning

Discrete diffusion and flow models are a promising alternative to autoregressive language models, but compressing many-step sampling into fewer steps typically requires distilling a pretrained teacher model. This caps the student at the teacher's quality and requires a costly two-stage training pipeline. We introduce Discrete Beckmann Transport Models (DBTM), built on a time-independent flow whose autonomous transport map provably carries any point in the ambient space to a fixed point on the vertices of the simplex in a single step. We show that this fixed-point property is characterized by a conservation equation whose residual can be minimized directly from data, removing the requirement for a teacher flow and time conditioning. Under this construction, a partially trained map corresponds to the flow truncated at finite time, so generation reduces to iterating one map until it reaches a fixed point. We further extend the map to a partial-context interpolant where additional function evaluations act as refinement steps rather than ODE integration steps. On language modeling and reasoning tasks, DBTM enables one- and few-step generation that improves quality and accuracy over discrete diffusion and continuous flow baselines.
Read the full abstract

Quantum Physics

Relative Entropy Decay via BKM coercivity for Quantum Markov Semigroups

We introduce the notion of the BKM coercivity constant and show that the positivity of this constant governs the exponential relative entropy decay of the semigroup. Indeed, we prove that the modified log-Sobolev constant is equivalent to the BKM coercivity constant up to a constant depending on the asymptotic conditional expectation. We also discover that the BKM coercivity constants for matrix amplifications are already attained with a qubit auxiliary system, which also coincides with the GNS spectral gap of the generating Lindbladian. As a corollary, we establish a criterion that the complete modified log-Sobolev inequality holds if and only if the GNS gap of the semigroup is strictly positive. All the discussion above applies to quantum Markov semigroups that admit a faithful asymptotic conditional expectation as long-time equilibration, with no detailed balance condition assumed. As an application, we show the finite-dimensional Chen-Kastoryano-Gilyén Gibbs samplers always satisfy the complete modified log-Sobolev inequality.
Read the full abstract

Optimization and Control

Strategic Index Reconstitution: Differential Games, Closed-Loop Equilibria and Mean-Field Dynamics

We study a continuous-time, multiasset trading game around index reconstitution, in which opportunistic traders have heterogeneous beliefs about the future index constitution and interact through transient cross-asset price impact and instantaneous execution costs. Opportunistic traders build inventory based on their individual beliefs about the future index constitution before a public announcement, adjust their positions to the revealed composition, and unwind them against the indexer's prescribed rebalancing trades at the implementation date when the new index composition takes effect. Under a no-price-manipulation condition, we construct a subgame-perfect Nash equilibrium. Its affine feedback policies are computed from a non-standard matrix Riccati equation and linear ODEs whose number and dimensions are independent of population size. Under mean-field scaling, we construct a mean-field equilibrium and obtain quantitative convergence and approximate-Nash bounds. Numerical illustrations show that greater competition among opportunists dampens price swings around implementation and that opportunistic trading lowers the indexer's implementation shortfall unless price resilience is very low. Holding the average belief fixed, greater belief dispersion can increase gross trading volume without changing net aggregate opportunist holdings or price impact.
Read the full abstract

High Energy Astrophysical Phenomena

Ce II-IV emission in kilonovae with R-matrix collision strengths and distorted wave recombination rates

We provide transition probabilities and Maxwellian-averaged collision-strengths for the electron-impact excitation of Ce ii - iv. These ions are potentially abundant in kilonovae (KNe) and feature both forbidden and allowed transitions in the infrared. These atomic data are obtained using the grasp$^0$ atomic structure code, interfaced with the non-perturbative darc R-matrix close-coupling codes. We additionally use autostructure to produce dielectronic and radiative recombination rate coefficients for Ce ii - vi. Using this data, we calculate an approximate ionization balance including non-thermal ionization rates with deposition rates from realistic models. We find Ce ii - iv to be of roughly equal abundance, subject to ejecta conditions. We find that generally we need $\gtrsim$ 10$^{-3}$ M$_{\odot}$ of Ce in total to produce emission of comparable luminosity to the features in the spectrum of AT2023vfi. In particular, it is found that [Ce iv] 4.5 $\mu$m requires both strong ionization and weak recombination, as well as a production of Ce considerably stronger than that of the solar abundance in order to have luminosity comparable to the apparent emission in this wavelength vicinity in \grb. %which is seemingly incompatible with the current understanding of the bulk composition of KNe. With such strict requirements on individual element masses, the detection of individual lanthanides on a line-by-line basis in the nebular phase seems unlikely - but lanthanides may remain important for cooling and early time analyses.
Read the full abstract

All authors Leo Patrick Mulholland, Niamh Ferguson, Luke J. Shingles, Catherine A. Ramsbottom, Connor P. Ballance, Stuart A. Sim

Statistical Methodology

Functional Estimation under Proxy-Based Full-Law Identification

We state general conditions under which the full-data law is identified in the presence of latent variables, leveraging key observed variables ("proxies") associated with unobserved variables. These assumptions extend those used in existing examples from the literature that recover the full-data law under relatively flexible model assumptions. We first describe graphical models compatible with our assumptions, and then consider target parameters that are functionals of the full-data law, developing estimating equations that give rise to consistent M-estimators using the observed data. Our approach replaces unobserved random variables appearing in full-data estimating equations with observed variables via a proxy-based weighting scheme. This strategy can be extended to construct observed-data influence functions, giving estimators with desirable properties such as multiple robustness and \(\sqrt{n}\)-consistency despite slower than parametric convergence of nuisance functions.
Read the full abstract

All authors Helen Guo, AmirEmad Ghassami, Ilya Shpitser, Elizabeth L. Ogburn

Chemical Physics

Toward Predictive Hydride Bond Energetics with Neural-Network Wavefunctions

Accurate prediction of transition-metal hydride (TM-H) bond dissociation energies (BDEs) remains challenging because of strong electron correlation, relativistic effects, and nuclear quantum contributions. Here, we assess the performance of neural-network variational Monte Carlo (NN-VMC) based on the Psiformer ansatz for systems with increasing complexity (LiH, OH, TiH and NiH), and compare it against CCSDT(Q)/CBS as well as available theoretical and experimental data. Throughout this study, both NN-VMC and ab initio calculations employ a common ccECP Hamiltonian to enable tractable and consistent comparisons. To reduce finite-training errors, we introduce complementary zero-variance and infinite-step extrapolation schemes. For LiH and OH systems, NN-VMC yields total energies that differ within sub-milli Hartree compared to CBS extrapolated ab initio results and the BDE differences remain under 2$\sigma$. In case of Ti and TiH, the variational NN-VMC energies at the end of training are already consistent with CCSD(T)/CBS, while post-training extrapolation systematically closes the gap towards the CCSDT(Q)/CBS results. In contrast, NiH provides a stringent test of wavefunction expressivity, where increasing the number of determinants in NN-wavefunction ansatz improves the recovered correlation energy. The comparison of NiH BDE reveals that, while the existing theoretical predictions cluster into distinct high and low BDE groups, the broad scatter in available experimental data prevents a definitive assessment of the most accurate theoretical approach. This work demonstrates that NN-VMC with ccECPs Hamiltonian provides a competitive framework for quantitative prediction of main-group and early TM-H energetics, while identifying late TM-H as an important benchmark for future developments in neural-network wavefunctions and electronic structure theory.
Read the full abstract

All authors Aqsa Shaikh, Lubos Mitas, P. Ganesh, Jaron T. Krogel

Statistical Mechanics

Bridging Control, Inference, Transport, and Thermodynamics: From Theory to Applications in Learning

The last decade has seen the development of powerful methods for learning complex structure from high-dimensional data. These advances have brought to the foreground fundamental connections between subdisciplines of physics, applied mathematics, and machine learning. In this review, we bring together some of these ideas, often expressed in different languages, to highlight a conceptual thread that links five distinct fields: control theory, optimal transport, probabilistic inference, non-equilibrium thermodynamics, and machine learning. A common theme is the optimization of free-energy-like functionals under dynamical or statistical constraints. We offer a guided tour through this thread and present selected applications in reinforcement learning, variational inference, and generative modeling. The review does not assume prior familiarity with these topics, and begins with principles originating from physics.
Read the full abstract

All authors Emmy Blumenthal, Nikolas Claussen, Benjamin Eysenbach, Catherine Ji, Gautam Reddy, Colin Scheibner, Benjamin Sorkin

High Energy Astrophysical Phenomena

Impact of uncertainties on the cosmic optical and infrared backgrounds on the propagation of astroparticles

When propagating through the universe, gamma rays at very-high energy (VHE, E > 100 GeV) and ultra-high energy cosmic rays (UHECRs, E > 1 EeV) can interact with the optical, infrared and microwave photon fields that permeate the universe. These interactions result in a characteristic absorption imprint in the spectra of extragalactic gamma-ray sources at VHE, and in a change in the mass-composition of UHECRs. The study of both VHE gamma rays and UHECRs therefore requires precise knowledge of the intensity of the cosmic photon fields. In this work, we explore the impact of the current uncertainties on the optical and infrared photon fields on the propagation of astroparticles. We restrict the range of available models to those best matching the recent measurements, and compare the different reconstructions resulting from these models. We find that the knowledge on the cosmic background light is no longer the dominant source of uncertainties in understanding the phenomenology of astroparticle sources in the low redshift universe (z < 0.1), enabling robust spectral and composition inference with the next generation of both gamma-ray and UHECR measurements.
Read the full abstract

All authors Lucas Gréaux, Antonio Condorelli, Leonel Morejon, Karl-Heinz Kampert, Jonathan Biteau

Robotics

Goal-Oriented Communications for Physical AI: Design and Testbed

Physical AI relies on frequently-updated, latency-sensitive video stream to perceive, reason, and interact with the physical world, resulting in strict latency requirements with much higher data volumes that existing 5G networks cannot support. Goal-oriented communication (GoC) offers as a promising approach to solve this challenge by transmitting only task-relevant semantic representations. However, existing GoC frameworks were mainly evaluated in the simulations while their effectiveness has never been validated in a practical deployment of physical AI application. In this work, we develop an end-to-end GoC testbed for Physical AI, which connects a PiPER robot arm equipped with an RGB-D camera and a 5G modem to an NVIDIA Jetson AGX Orin edge server through a 5G OpenAirInterface network. We propose and implement three GoC frameworks that transmit 3D bounding boxes, 2D scene graphs, and 3D scene graphs, as three types of semantic representations, respectively. They share the common functional modules designed for closed-loop Physical AI applications, including semantic extraction, full stack 5G transmission, language model inference, digital twin validation, and robotic control. Extensive experiments on our testbed show that our GoC frameworks reduce the task completion time by up to 52.6% and improve task success probability by up to 45%, compared to the traditional framework that periodically transmits the raw image data. These results validate the practical effectiveness of our GoC framework and pave the way for efficient and reliable Physical AI applications over future 6G networks. Project website: https://sites.google.com/view/goc-physical-ai-testbed.
Read the full abstract

All authors Shutong Chen, Wenkai Zhang, Adnan Aijaz, Miao Guo, Yansha Deng

Machine Learning (Statistics)

Quenched Ensemble Sampling

Some of the sharpest challenges in sampling from the energy functions of physical systems arise at phase transitions, where the density of states changes abruptly and many sampling algorithms stall. Nested sampling is a particle method that traverses the density of states under a hard energy constraint and is known to be robust to such transitions, but its application in high dimension is limited by the difficulty of sampling under that constraint. In this work we introduce Quenched Ensemble Sampling, which generalises the hard constraint to a family of repulsive potentials at the energy boundary. This preserves the quenched path of monotonically decreasing energy while making the constrained target amenable to scalable gradient-based kernels. We demonstrate on synthetic models of phase transitions that our method estimates the marginal likelihood and draws posterior samples across a first-order transition where popular alternatives such as tempering fail. We apply the procedure to marginal likelihood estimation in Bayesian neural networks, enabling model comparison between network architectures. Finally, in a high-dimensional continuous lattice field theory, we show that this method traverses a first-order transition and estimates the partition function.
Read the full abstract

Combinatorics

A short proof of the Erd\H os–Sós Conjecture

The Erd\H os–Sós Conjecture was recently proved by GPT-6 Astra, using a very ingenious and surprising argument. In this note, we present a simplified version of this argument in an (arguably) more natural form. We also prove a related conjecture of Addario-Berry, Havet, Linhares Sales, Reed and Thomassé, and give a substantially simplified proof of an analogous result for eulerian digraphs, again proved by GPT-6 Astra, and announced by Mubayi and Verstraete. In addition, we determine the extremal (di)graphs for all three problems.
Read the full abstract

Algebraic Geometry

Geometric realizations of Brauer classes on K3 surfaces from hyperkähler contractions

Elements of the Brauer group $\operatorname{Br}(S)$ of a variety $S$ have geometric incarnations as étale-projective $S$-bundles, yet producing minimalist constructions of such bundles, which often power arithmetic applications, remains a difficult problem. When $S$ is a K3 surface with Picard rank $1$, we use the birational geometry of moduli spaces of twisted sheaves on $S$ to construct geometric realizations of nontrivial elements of $\operatorname{Br}(S)$. We recover many known geometric constructions of Brauer classes on K3 surfaces while providing a common moduli-theoretic framework for them. As a by-product, we give a new proof of the period-index theorem for very general K3 surfaces.
Read the full abstract

Statistical Mechanics

Modulated structures in discrete approximations to a three-dimensional cholesteric model on a recursive lattice

We study finite-state approximations to a chiral nematic lattice model whose local directors are embedded in three-dimensional orientational space. The model analyzed here is formulated in terms of a rotated quadrupolar interaction, which provides a direct angular chirality parameter $\Delta$ and reduces to the usual planar chiral-nematic form for directors confined to the xy plane. On a rooted Cayley tree, the statistical problem is written as a nonlinear recursion for branch partition functions. In the infinite-coordination limit, with $rJ_Q$ fixed, the recursion becomes a softmax map for the state probabilities. We analyze Cartesian three-state, planar four-state, and non-planar four-state discretizations. The first supports ordered and period-two attractors, whereas the planar four-state discretization displays a sequence of commensurate and longer-period modulated attractors. Its disordered fixed point loses stability at $t=9/16$ through a complex-conjugate pair with phase $q=\pm 2\Delta$, so a modulated critical mode is selected for every nonzero chirality. The non-planar four-state set is intrinsically anisotropic and, when the stated model is reconstructed directly, supports prominent period-three as well as period-two and higher-period cycles. Because attractor stability does not by itself establish global thermodynamic stability, the diagrams reported here are interpreted as attractor/stability diagrams rather than equilibrium phase diagrams.
Read the full abstract

Geometric Topology

PL Recognition After Two $S^2\times S^2$ Stabilisations is PSPACE-Hard

Let $Z$ be any fixed closed connected PL $4$-manifold. We give a polynomial-time many-to-one reduction from the compressed word problem in Thompson's group $F$ to fixed-target recognition of $Z\#_{2}(S^2\times S^2)$. Once a finite triangulation of $Z$ has been fixed, the construction sends a straight-line program $A$ to a closed triangulated PL $4$-manifold $X_{A,Z}$ such that $\operatorname{val}(A)=1$ in $F$ if and only if $X_{A,Z}$ is PL-homeomorphic to $Z\#_2(S^2\times S^2)$. Since the compressed word problem in $F$ is PSPACE-complete, every such recognition problem is PSPACE-hard. In particular, fixed-target recognition of $\#_n(S^2\times S^2)$ is PSPACE-hard for every fixed $n\geq2$.
Read the full abstract

Distributed Computing

Cnuas: A Software-Defined AI/HPC Rack-scale Emulation Platform and Hyperscale Data Center Facility Twin

Modern AI and HPC systems integrate accelerators, high-speed networks, and management controllers at rack scale. Developing software for this infrastructure typically requires access to scarce, costly hardware, while software abstractions can obscure how workloads depend on resources across servers and accelerators. This paper presents Cnuas, an open-source, experimental rack-scale emulation platform whose baseline architecture follows the Open Compute Project (OCP) Open Rack v3 specifications. Through functional emulation, it supports experimentation, learning and software development within academic and industrial research and development, rather than matching the throughput or latency of physical hardware. Its web-based user interface visualizes racks, devices and their interconnections to help developers build a system-level mental model of the infrastructure supporting their workloads. At its core, CnuasNIC and CnuasSwitch implement a guest-visible remote direct memory access (RDMA) adapter and a host-resident hybrid software switch supporting both RoCEv2 and native InfiniBand. The platform also provides a dedicated AI/ML accelerator (GPU) peer fabric and OpenBMC-based rack management with executable power supply and battery backup firmware over RS-485. These components support the study of device, driver, and firmware interfaces on commodity hosts. The accelerator software stack remains an early research prototype, and facility modeling with OpenUSD is an exploratory extension. The paper presents the architecture, interfaces, and bounded prototype results as a basis for community collaboration across the core platform and its extensions.
Read the full abstract