https://arxiv.org/api/Zjv8Nq0FVwm7/pRgxXprQ6F9qsw 2026-06-21T13:24:55Z 8189 600 15 http://arxiv.org/abs/2508.12314v1 2025-08-17T10:16:41Z

We present a novel interdisciplinary framework that bridges synchronization theory and multi-agent AI systems by adapting the Kuramoto model to describe the collective dynamics of heterogeneous AI agents engaged in complex task execution. By representing AI agents as coupled oscillators with both phase and amplitude dynamics, our model captures essential aspects of agent specialization, influence, and communication within networked systems. We introduce an order parameter to quantify the degree of coordination and synchronization, providing insights into how coupling strength, agent diversity, and network topology impact emergent collective behavior. Furthermore, we formalize a detailed correspondence between Chain-of-Thought prompting in AI reasoning and synchronization phenomena, unifying human-like iterative problem solving with emergent group intelligence. Through extensive simulations on all-to-all and deterministic scale-free networks, we demonstrate that increased coupling promotes robust synchronization despite heterogeneous agent capabilities, reflecting realistic collaborative AI scenarios. Our physics-informed approach establishes a rigorous mathematical foundation for designing, analyzing, and optimizing scalable, adaptive, and interpretable multi-agent AI systems. This work opens pathways for principled orchestration of agentic AI and lays the groundwork for future incorporation of learning dynamics and adaptive network architectures to further enhance system resilience and efficiency.

2025-08-17T10:16:41Z 9 pages, 6 figures Chiranjit Mitra http://arxiv.org/abs/2506.04275v2 2025-08-17T09:03:59Z

Self-organized criticality is a well-established phenomenon, where a system dynamically tunes its structure to operate on the verge of a phase transition. Here, we show that the dynamics inside the self-organized critical state are fundamentally far more versatile than previously recognized, to the extent that a system can self-organize to a new type of phase transition while staying on the verge of another. In this first demonstration of self-organization to multicriticality, we investigate a model of coupled oscillators on a random network, where the network topology evolves in response to the oscillator dynamics. We show that the system first self-organizes to the onset of oscillations, after which it drifts to the onset of pattern formation while still remaining at the onset of oscillations, thus becoming critical in two different ways at once. The observed evolution to multicriticality is robust generic behavior that we expect to be widespread in self-organizing systems. Overall, these results offer a unifying framework for studying systems, such as the brain, where multiple phase transitions may be relevant for proper functioning.

2025-06-04T00:30:30Z Silja Sormunen Thilo Gross Jari Saramäki http://arxiv.org/abs/2505.11437v3 2025-08-15T19:14:22Z

We investigate the role of connection density in an adaptive network model of chaotic units that dynamically rewire based on their internal states and local coherence. By systematically varying the network's connectivity density, we uncover distinct dynamical regimes and structural transitions, revealing mechanisms of spontaneous modularity, dynamical segregation, and integration. We find that at higher densities, the network exhibits both local clustering and global synchrony. Additionally, we observe that low-density networks tend to fragment into desynchronized clusters, while high-density networks converge to synchronized states combining strong global integration with persistent modular segregation. Inspired by neural architectures, our model provides a general framework for understanding how simple microscopic rules can give rise to complex emergent behaviors in dynamical networks.

2025-05-16T16:57:17Z Ramiro Plüss Pablo Martín Gleiser http://arxiv.org/abs/2508.10818v1 2025-08-14T16:44:21Z

Limit cycles are self-sustained, closed trajectories in phase space representing (un)-stable, periodic behavior in nonlinear dynamical systems. They underpin diverse natural phenomena, from neuronal firing patterns to engineering oscillations. The presence of multiple concentric limit cycles reflects distinct behavioral symmetries within a system. In this work, we investigate the hierarchical dynamical transitions from one limit cycle to another, driven by oscillatory excitation while preserving other system properties. We demonstrate that controlling multirhythmicity through hierarchical, stepwise periodic modulation enables reliable switching between rhythmic states. This hierarchical control framework is crucial for applications in neuro-engineering and synthetic biology, where precise, robust modulation of complex rhythmic behaviors enhances system functionality and adaptability.

2025-08-14T16:44:21Z 10 pages, 6 figures Sandip Saha Suvam Pal Dibakar Ghosh http://arxiv.org/abs/2508.09827v1 2025-08-13T14:03:10Z

Silicon is the most important semiconductor electrode with applications in photoelectrochemistry and sensor technology. Yet its electrochemistry exhibits many poorly understood phenomena, including oscillations during the anodic dissolution of silicon electrodes. In this article, we present a mathematical model based on physicochemical steps that captures these oscillations and enables a thorough understanding of the underlying mechanism. The model describes the formation and dissolution of an oxide layer, and determines the oxide composition and the electrostatic potential in the direction perpendicular to the electrode. Oscillations occur if the following conditions are fulfilled: 1. The etching speed increases with defects in the oxide layer 2. The number of defects decreases with increasing electric field strength at the Si-oxide interface. 3. There is a sufficient time delay between the production and the etching of the oxide. Numerical simulations reproduce experimental results well, including the dependence of the oscillation amplitude and period on the potential, as well as the hysteresis behavior observed in cyclic voltammetry. Based on these results, we derive a simplified time-delay differential equation model. Using linear stability analysis, we confirm the essential role of the time delay for the oscillatory oxide-layer dynamics. The basic steps of the model are general and in line with the point defect model for growth and dissolution of passive films on metal electrodes. Therefore, it is likely applicable to a variety of oscillating anodic metal or semiconductor dissolution reactions.

2025-08-13T14:03:10Z 16 pages, 17 figures Yukiteru Murakami Katharina Krischer http://arxiv.org/abs/2508.09618v1 2025-08-13T08:49:59Z

Training highly recurrent networks in continuous action spaces is a technical challenge: gradient-based methods suffer from exploding or vanishing gradients, while purely evolutionary searches converge slowly in high-dimensional weight spaces. We introduce a hybrid, derivative-free optimization framework that implements reinforcement learning by coupling global evolutionary exploration with local direct search exploitation. The method, termed ENOMAD (Evolutionary Nonlinear Optimization with Mesh Adaptive Direct search), is benchmarked on a suite of food-foraging tasks instantiated in the fully mapped neural connectome of the nematode \emph{Caenorhabditis elegans}. Crucially, ENOMAD leverages biologically derived weight priors, letting it refine--rather than rebuild--the organism's native circuitry. Two algorithmic variants of the method are introduced, which lead to either small distributed adjustments of many weights, or larger changes on a limited number of weights. Both variants significantly exceed the performance of the untrained connectome (in what can be interpreted as an example of transfer learning) and of existing training strategies. These findings demonstrate that integrating evolutionary search with nonlinear optimization provides an efficient, biologically grounded strategy for specializing natural recurrent networks towards a specified set of tasks.

2025-08-13T08:49:59Z 15 pages, 5 figures Miles Walter Churchland Jordi Garcia-Ojalvo http://arxiv.org/abs/2507.07282v3 2025-08-12T16:43:28Z

The overdamped Josephson junction in superconductivity theory can be modeled by the family of dynamical systems on the torus, which is known as the RSJ model. This family admits an equivalent description by a family of second-order differential equations: special double confluent Heun equations. In the present paper, we construct two new families of dynamical systems on torus that can be equivalently described by a family of general Heun equations (GHE), with four singular points, and confluent Heun equations, with three singular points. The first family, related to GHE, is a deformation of the RSJ model, which will be denoted by dRSJ. The phase-lock areas of a family of dynamical systems on the torus are those level subsets of the rotation number function that have nonempty interiors. It is known that for the RSJ model, the rotation number quantization effect occurs: phase-lock areas exist only for integer rotation number values. Moreover, each phase-lock area is a chain of domains separated by points. Those separation points that do not lie on the abscissa axis are called constrictions. In the present paper, we study phase-lock areas in the new family dRSJ. The quantization effect remains valid in this family. On the other hand, we show that in the new family dRSJ the constrictions break down.

2025-07-09T20:59:29Z Artem Alexandrov Alexey Glutsyuk http://arxiv.org/abs/2508.10049v1 2025-08-12T10:23:16Z

We propose a universal physical mechanism for the emergence of 1/f fluctuations, observed across a wide range of systems. In particular, we verify this on acoustic cases. The mechanism is based on amplitude modulation (AM) and demodulation (DM), where the 1/f spectral law arises not in the raw waveform but in its demodulated amplitude envelope. Two distinct yet complementary processes generate the required AM: (i) stochastic synchronization among oscillators, modeled via an extended Kuramoto framework that captures perpetual synchronization-desynchronization cycles, and (ii) frequency-selective resonance, modeled by spectral accumulation of eigenmodes in acoustic or structural environments. Numerical simulations demonstrate that both mechanisms, acting separately or in combination, robustly produce 1/f spectra over several decades when DM is applied, and that the classical Kuramoto critical point is not necessary for their emergence. We demonstrate the cross-domain relevance of this AM/DM framework through analyses of musical performances, seismic records, and astrophysical time series, revealing a common underlying structure. This work establishes demodulation as a general route to 1/f fluctuations, providing a simple and scalable explanation for its ubiquity in both natural and engineered systems. Keywords: 1/f fluctuation, amplitude modulation, synchronization, resonance, Kuramoto model, music, natural noise, demodulation

2025-08-12T10:23:16Z 14 pages, 10 figures Akika Nakamichi Izumi Uesaka Masahiro Morikawa http://arxiv.org/abs/2403.19297v2 2025-08-11T04:22:05Z

We have recently proposed a fully quantum-mechanical definition of the asymptotic phase for quantum nonlinear oscillators, which is also applicable in the strong quantum regime [Kato and Nakao 2022 Chaos 32 063133]. In this study, we propose a definition of the quantum asymptotic amplitude for quantum oscillatory systems, which naturally extends the asymptotic amplitude for classical nonlinear oscillators on the basis of the Koopman operator theory. We introduce the asymptotic amplitude for quantum oscillatory systems by using the eigenoperator of the backward Liouville operator associated with the largest non-zero real eigenvalue. Using examples of the quantum van der Pol oscillator with the quantum Kerr effect, which exhibits quantum limit-cycle oscillations, and the quantum van der Pol model with the quantum squeezing and degenerate parametric oscillator with nonlinear damping, which exhibit quantum noise-induced oscillations, we demonstrate that the proposed quantum asymptotic amplitude yields isostable amplitude values that decay exponentially with a constant rate. Furthermore, using the quantum asymptotic amplitude, we introduce effective quantum periodic orbits for quantum limit-cycle oscillations and quantum noise-induced oscillations.

2024-03-28T10:33:29Z 31 pages, 6 figures. Updated version. Reducing text overlap with arXiv:2302.05584 Yuzuru Kato http://arxiv.org/abs/2305.16659v3 2025-08-08T17:51:20Z

Active matter composed of self-propelled particles features fascinating self-organization phenomena, spanning from motility-induced phase separation to phototaxis to topological excitations depending on the nature and parameters of the system. In the present paper, we consider micelle formation by active particles with a broken symmetry having a circular back and a sharpened nose toward which the particles accelerate. As we demonstrate in experiments with robotic swarms, such particles can either remain in the isotropic phase or form micelles depending on the location of their center of inertia, in accordance with a recent theoretical proposal [T. Kruglov and A. Borisov, Presentations and Videos to 7th Edition of the International Conference on Particle-based Methods (2021), Vol. CT07, p. 2]. Such a behavior is observed for both nonchiral particles moving linearly and placed in a parabolic potential and for chiral particles moving along circular trajectories on a flat surface. By performing experiments with single robots and two-robot collisions, we unveil that the observed emergence of micellization associated with shifting robots' center of inertia towards their noses is governed by at least two-particle effects, in particular, by a difference in the formation of stable two-robot clusters. Finally, we consider the dependence of micelle lifetime and formation probability as well as two-robot collisions on friction between the lateral surfaces of the robots. Crucially, the predicted micellization does not involve any solvation shells that give rise to the micellization of surfactants but is instead driven by an interplay of activity and particle shape asymmetry.

2023-05-26T06:08:27Z 18 pages, 8 figures + Supplemental Material Phys. Rev. E 111, 065424 (2025) Anastasia A. Molodtsova Mikhail K. Buzakov Oleg I. Burmistrov Alina D. Rozenblit Vyacheslav A. Smirnov Daria V. Sennikova Vadim A. Porvatov Ekaterina M. Puhtina Alexey A. Dmitriev Nikita A. Olekhno 10.1103/4yzc-ppgy http://arxiv.org/abs/2507.03960v2 2025-08-08T04:33:41Z

A longstanding puzzle in urban science is whether there's an intrinsic match between human populations and the mass of their built environments. Previous findings have revealed various urban properties scaling nonlinearly with population, yet existing models of city built mass are still dominated by per-capita linear thinking. Our analysis of >3,000 cities globally reveals universal sublinear scaling of city mass with population at both the city (\{beta}=0.90) and neighborhood levels (δ=0.75). This means that larger cities and denser neighborhoods achieve economies of scale with less per-capita built mass. Our theoretical framework further shows that city-level scaling emerges naturally from within-city disparities. This multi-scale understanding redefines "over-built" and "under-built" conditions as deviations from expected scaling patterns, implying either excessive environmental impacts or inadequate living standards. Effective urban policy thus requires moving beyond simple per-capita assumptions, adopting scale-adjusted metrics and managing cities as nested, complex systems.

2025-07-05T08:53:10Z 4 main figures Kangning Huang Mingzhen Lu http://arxiv.org/abs/2508.05766v1 2025-08-07T18:28:54Z

This paper proposes a novel framework for developing safe Artificial General Intelligence (AGI) by combining Active Inference principles with Large Language Models (LLMs). We argue that traditional approaches to AI safety, focused on post-hoc interpretability and reward engineering, have fundamental limitations. We present an architecture where safety guarantees are integrated into the system's core design through transparent belief representations and hierarchical value alignment. Our framework leverages natural language as a medium for representing and manipulating beliefs, enabling direct human oversight while maintaining computational tractability. The architecture implements a multi-agent system where agents self-organize according to Active Inference principles, with preferences and safety constraints flowing through hierarchical Markov blankets. We outline specific mechanisms for ensuring safety, including: (1) explicit separation of beliefs and preferences in natural language, (2) bounded rationality through resource-aware free energy minimization, and (3) compositional safety through modular agent structures. The paper concludes with a research agenda centered on the Abstraction and Reasoning Corpus (ARC) benchmark, proposing experiments to validate our framework's safety properties. Our approach offers a path toward AGI development that is inherently safer, rather than retrofitted with safety measures.

2025-08-07T18:28:54Z Bo Wen http://arxiv.org/abs/2508.05619v1 2025-08-07T17:57:12Z

This paper argues that Active Inference (AIF) provides a crucial foundation for developing autonomous AI agents capable of learning from experience without continuous human reward engineering. As AI systems begin to exhaust high-quality training data and rely on increasingly large human workforces for reward design, the current paradigm faces significant scalability challenges that could impede progress toward genuinely autonomous intelligence. The proposal for an ``Era of Experience,'' where agents learn from self-generated data, is a promising step forward. However, this vision still depends on extensive human engineering of reward functions, effectively shifting the bottleneck from data curation to reward curation. This highlights what we identify as the \textbf{grounded-agency gap}: the inability of contemporary AI systems to autonomously formulate, adapt, and pursue objectives in response to changing circumstances. We propose that AIF can bridge this gap by replacing external reward signals with an intrinsic drive to minimize free energy, allowing agents to naturally balance exploration and exploitation through a unified Bayesian objective. By integrating Large Language Models as generative world models with AIF's principled decision-making framework, we can create agents that learn efficiently from experience while remaining aligned with human values. This synthesis offers a compelling path toward AI systems that can develop autonomously while adhering to both computational and physical constraints.

2025-08-07T17:57:12Z Bo Wen http://arxiv.org/abs/2508.08301v1 2025-08-07T04:37:16Z

Humans are bounded rational at best and this, we argue, has worked in their favour in the hunter-gatherer society where emergence of a coordinated action, leading to cooperation, is otherwise the standard stag-hunt dilemma (when individuals are rational). In line with the fact the humans strive for developing self-reputation by having less propensity to cheat than to be cheated, we observe that the payoff structure of the stag-hunt game appropriately modifies to that of coordination-II game. Subsequently, within the paradigm of evolutionary game theory, we establish that a population -- consisting of procedural rational players (a type of bounded rationality) -- is unequivocally evolutionarily stable against emergence of more rational strategies in coordination-II game. The cooperation is, thus, shown to have been established by evolutionary forces picking less rational individuals.

2025-08-07T04:37:16Z J. Phys. Complex. 6, 035004 (2025) Joy Das Bairagya Sagar Chakraborty 10.1088/2632-072X/adf2ee http://arxiv.org/abs/2506.18699v2 2025-08-06T09:21:12Z

Many examples from quantum and classical physics are known where topological protection is responsible for the robustness of the dynamics. Less explored is the role of topological protection in the context of classical oscillatory systems. As on-site dynamics, we consider prototypical oscillator models with possible applications in biochemical systems. However, our choice of coupling geometry is inspired by models from condensed matter physics which -- in isolation -- guarantee non-trivial topology in momentum space. We choose directed couplings between units on a two-dimensional grid, alternating between weak and strong values, such that oscillations become localized at the edges of the grid while bulk units transition to oscillation-death states, resulting in a frequency chimera-like state. These patterns are resilient to parameter mismatches, additive noise, and structural defects. To explain the robustness of these edge oscillations we use topological characteristics and calculate Zak phases. As it turns out, the edge-localized oscillations result from a bulk-boundary correspondence that applies to our system even though our derived effective Hamiltonian is non-Hermitian. By appropriately tuning the system parameters, it is possible to control which regions of the two-dimensional grid are in an oscillatory state and which settle to oscillation-death states.

2025-06-23T14:37:00Z 24 Pages, 21 Figures Sayantan Nag Chowdhury Hildegard Meyer-Ortmanns