https://arxiv.org/api/xrNIcva5s65TsgeA0G7aAhuX1DA 2026-06-10T02:51:32Z 26559 30 15 http://arxiv.org/abs/2406.05335v3 2026-06-06T04:54:38Z

Generation of text and speech in natural languages can be modeled as a stochastic process. This idea dates back to the seminal work of Markov and, later, to that of Shannon and also underlies the recent development of large language models (LLMs). The stochastic processes corresponding to natural languages should be distinct from those that generate nonlinguistic sequences. One of the features that discriminate linguistic and nonlinguistic sequences is power-law behavior, which is universally observed across different languages. In statistical physics, such behavior suggests that natural languages are critical: They lie near a phase transition point in a parametrized space of stochastic processes. However, testing this conjecture is not straightforward. A phase transition, even if it exists, cannot be directly observed in real-world natural languages because they do not have any controllable parameters. Here, we use LLMs as controllable effective models of natural languages. Through statistical analyses of texts generated by LLMs, we find that, when a parameter analogous to physical temperature is varied, LLMs undergo a phase transition. The transition separates a low-temperature phase with complex repetitive structures in generated texts from a high-temperature phase in which LLMs generate incomprehensible texts. At the critical point between these phases, generated texts display the power-law behavior similar to that of natural languages and most closely resemble natural languages as measured by a standard metric in natural language processing. These findings strongly suggest that natural languages are indeed critical.

2024-06-08T03:37:05Z 8 pages, 6 figures Kai Nakaishi Yoshihiko Nishikawa Koji Hukushima http://arxiv.org/abs/2602.02648v2 2026-06-06T01:16:11Z

Quantum criticality in the presence of strong quenched randomness remains a challenging topic in modern condensed matter theory. We show that the topology and anomaly associated with average symmetry can be used to predict certain nontrivial universal properties. Our focus is on systems subject to average Lieb--Schultz--Mattis constraints, where lattice translation symmetry is preserved only on average, while on-site symmetries remain exact. We argue that in the absence of spontaneous symmetry breaking and intrinsic topological order, the system must exhibit critical correlations of local operators in two distinct ways: (i) for some operator $O_e$ charged under exact symmetries, the first absolute moment correlation $\overline{|\langle O_e(x)O^{\dagger}_e(y)\rangle|}$ decays slowly; and (ii) for some operator $O_a$ charged under average symmetries, the first-moment correlation $\overline{\langle O_a(x)O^{\dagger}_a(y)\rangle}$ decays slowly. We verify these predictions in a few examples: the random-singlet Heisenberg spin chain in one dimension, and the disordered free-fermion critical states in symmetry class BDI in one and two dimensions. Surprisingly, even for these well-studied systems, our anomaly-based argument reveals critical correlations overlooked in previous literature. We also discuss the experimental feasibility of measuring these critical correlations.

2026-02-02T19:00:01Z 6+13 pages, 1+7 figures, A few clarifications added Yasamin Panahi Subhayan Sahu Naren Manjunath Chong Wang http://arxiv.org/abs/2606.07774v1 2026-06-05T18:31:26Z

For the disordered Ising chain with transverse and longitudinal fields, where couplings and fields are drawn from strictly positive distributions, Cao~\cite{Cao} has shown that the moments $μ_{2k} = \|[H,σ^z_0]^{(k)}\|_2^2$ grow almost factorially, $μ_{2k}^{1/(2k)}\sim k/\ln k$, and thus asymptotically at the maximal allowed rate. We generalize this result by resolving the operator norm in support length and show that the weight at length $\ell_k \sim k/\ln k$ already exhibits almost factorial growth, $\|[H,σ^z_0]^{(k)}_{\ell_k}\|_2 \gtrsim (k/\ln k)^k$. This implies maximal spatial delocalization of local operators and, in particular, rules out dynamical locality -- the strongest form of many-body localization -- at any disorder strength. We further establish rigorously a finite-size crossover scale $L\sim (W/J)^2$, where $W$ is the disorder and $J$ the coupling strength. For $L\lesssim (W/J)^2$ numerical studies only access a pre-asymptotic regime. Finally, we identify a structural path-entropy obstruction to perturbative LIOM constructions, based on the almost factorial branching of operator content and independent of resonance effects; the same mechanism strongly suggests ballistic real-time operator spreading, so sub-ballistic or localized dynamics would require a presently unidentified cancellation principle acting on almost factorially many disorder-dependent paths with random amplitudes.

2026-06-05T18:31:26Z J. Sirker http://arxiv.org/abs/2606.07331v1 2026-06-05T14:46:54Z

Ising machines are a promising approach to solve combinatorial optimization problems. They map these problems onto the Ising model and search for low-energy configurations. However, navigating the rugged energy landscapes of these systems remains difficult. To improve this navigation, classical adiabatic annealing has been proposed in the literature as a heuristic optimization method for classical Ising machines. Using this technique, the Hamiltonian of the Ising machine is gradually transformed from an easily solvable Hamiltonian to the target Hamiltonian. However, its purported effectiveness is primarily motivated by an analogy to quantum adiabatic annealing, and systematic benchmarking has remained limited. In this work, we analyze the classical adiabatic annealing technique using continuation methods. Motivated by insights from this analysis, we propose an optimized annealing strategy we refer to as hybrid classical adiabatic annealing. We benchmark our proposed strategy using MaxCut instances with up to 800 spins and problems with external fields, for which it achieves a marginal improvement for a limited set of problems. We conclude that, although theoretically motivated and occasionally beneficial, the hybrid strategy does not offer a sufficient practical advantage over simpler, existing techniques.

2026-06-05T14:46:54Z 14 pages, 10 figures and 2 tables Jacob Lamers Guy Verschaffelt Guy Van der Sande http://arxiv.org/abs/2606.07327v1 2026-06-05T14:45:06Z

Machine-learned interatomic potentials (MLIPs) have had a profound impact on molecular modelling in recent years, promising to resolve the long-standing tension between the scale and accuracy of simulations. There has been a proliferation of new models and designs, and recently the paradigm of ``foundational'' MLIPs has become prevalent. Broadly speaking, foundation models are trained on large diverse datasets and promise to work well for new systems with minimal updates required. However, in such a new and fast moving field, there are many unanswered questions. In this article, we set out to articulate and explore what we see as the most important among these questions. We start by developing a working definition for foundational MLIPs and use this definition to frame the subsequent open questions. Despite the rapid progress in the field of MLIP models, we believe that these are fundamental questions which will continue to define cutting edge research in MLIPs in the years to come.

2026-06-05T14:45:06Z Isabel Creed Tim Rein Ingvars Vitenburgs Wojciech G. Stark Viktor Ellingsson Ahmed Y. Ismail Guangyu Liu Yuchen Lou Bradley A. A. Martin Cyprien Bone Matthew A. H. Walker Mueen Taj Shirui Wang Kelvin Wong Ruiqi Wu Prakriti Kayastha Bingqing Cheng Aditi Krishnapriyan Michele Ceriotti Marcel F. Langer Jarvist Moore Frost Alex M. Ganose Venkat Kapil Keith T. Butler http://arxiv.org/abs/2606.07274v1 2026-06-05T13:50:19Z

We study a one-dimensional long-range Su-Schrieffer-Heeger model with third-nearest-neighbor hopping and subject to quasiperiodic disorder. In the clean limit, the model hosts phases characterized by winding numbers $W=-1,0,1$ and $2$. The introduction of quasiperiodic disorder profoundly modifies the phase diagram and induces a series of topological phase transitions. Owing to the competition between topological dimerization and localization, topological Anderson insulating (TAI) phases with different winding numbers emerge and can persist even when the spectral gap becomes nearly closed in the strong-disorder regime. In addition, we uncover multiple reentrant topological phase transitions induced by varying either the quasiperiodic disorder strength or the hopping amplitudes. Remarkably, the system exhibits staircase-like topological Anderson transitions, where the real-space winding number evolves through successive quantized steps with increasing disorder strength. Our results demonstrate that the interplay between long-range hopping and quasiperiodic disorder generates a rich landscape of disorder-induced topological phases and reentrant topological transition phenomena.

2026-06-05T13:50:19Z 9 pages, 6 figures Fang-Ming Meng Qi-Bo Zeng http://arxiv.org/abs/2606.07268v1 2026-06-05T13:42:28Z

We investigate the problem of the time reversal symmetry of fluctuations, as witnessed by the average shape of avalanches. This quantity has been measured in a variety of systems, ranging from magnetic materials to earthquakes. Although an asymmetric shape is often observed, which is a signature of a non-equilibrium dynamics, there is no general theoretical control of this feature. In this paper, we propose a non equilibrium extension of a paradigmatic model for ``crackling-noise'', the so called ABBM model. Our model is strictly related to the Brownian Gyrator, which has been previously introduced in stochastic thermodynamics as the simplest model for thermal anisotropy, but it can also be framed in the context of rate-and-state models. It reproduces the phenomenology observed in experiments on granular friction, and allows for a systematic theoretical study of the asymmetry. We manage to correlate a measure of asymmetry, that can be easily computed in experiments, with the entropy production rates of the dynamics.

2026-06-05T13:42:28Z 9 pages, 6 figures + SI (2 pages, 1 figure) Andrea Baldassarri Andrea Puglisi http://arxiv.org/abs/2606.07262v1 2026-06-05T13:38:57Z

Structures that capture functionality in the form of animate or intelligent machines have the potential to transform modern engineering applications. Animation and embedded intelligence are typically realised by integrating advanced capabilities such as reversibility, adaptive responses and learning directly into the materials themselves. Currently, the majority of adaptive material systems rely on predefined adaptive designs combined with in-service, electronics-based computing to dynamically modify the structural behaviour. However, structural configurations with interconnected adaptable nodes are able to approximate continuous functions, providing new possibilities and opportunities than classical metamaterials and computational materials. We discuss here the potential to design load-bearing engineering materials with trainable physical parameters and neural network-inspired morphologies, embedding intelligence directly into their structure, a concept we define as Engineering Material Neural Networks (EMNNs) as a subcategory of Physical Neural Networks. In this perspective, we first establish the foundational concept of EMNNs; we then detail the mechanical and multifunctional properties required for such structural configurations. Finally, we evaluate existing and emerging engineering materials that hold promise for enabling this innovative approach. Key material candidates for realising EMNNs include composites, architected, biological and engineering living materials. We also outline future directions in materials science and structural engineering for developing EMNNs.

2026-06-05T13:38:57Z 37 pages, 8 pages Charles de Kergariou Hortense le Ferrand Ali Momeni Romain Fleury Kunal Masania Adam W Perriman Fabrizio Scarpa http://arxiv.org/abs/2606.07247v1 2026-06-05T13:17:18Z

Neural ordinary differential equations (neural ODEs) have rapidly gained prominence as a powerful and unifying framework for conceptualizing artificial neural networks, elegantly connecting the continuous-time modeling of dynamical systems with the discrete, data-driven paradigm of modern deep learning. Beyond their practical advantages they offer fresh theoretical insights into the training and generalization properties of neural networks. The distinctive feature of this framework is its dual dynamical nature: inference dynamics, which govern the ODE evolution during forward computation, and training dynamics, which control the optimization of model parameters. This makes neural ODEs a particularly well-suited theoretical framework for studying a large variety of settings such as multi-layer neural networks (ResNets for example), autoregressive models (with next-token generation dynamics), generative models, and recurrent neural networks in theoretical neuroscience. In this work, we introduce a theoretically grounded class of models for studying neural ODEs trained via online stochastic gradient descent. We solve the training dynamics of these models via dynamical mean field theory and derive learning curves in the high-dimensional limit.

2026-06-05T13:17:18Z 28 pages, 2 figures Pierfrancesco Urbani http://arxiv.org/abs/2606.07241v1 2026-06-05T13:09:28Z

In this paper, the class of randomised mixed labyrinth fractals is introduced. It is a class of finitely ramified Sierpinski carpets that generalize mixed labyrinth fractals. The structures are generated by randomly selected labyrinth patterns with fixed selection probabilities at each iteration level, offering a flexible framework to study fractal topology, arc dimensions, and shortest path properties. Here, the focus lies on analysing how the randomised mixing of patterns - specifically their shape, symmetry, and path geometry - effects arc dimensions, path lengths, and isotropy restoration. The study reveals that isotropy, previously shown for self-similar fractals, extends to the randomised mixed class. Various scaling behaviours of shortest path dimensions with respect to the mixing probability are identified, including linear and nonlinear monotonic trends, as well as transitions with maxima. The approximated path matrix is proposed as an efficient alternative to extensive iterative simulations, reliably reproducing statistical results. The findings highlight the relevance of pattern properties in determining fractal structures and dynamics and suggest applications in physical systems such as diffusion, signal processing, and antenna design.

2026-06-05T13:09:28Z 25 pages, 16 figures, 3 tables Janett Prehl Ligia Loretta Cirstea Daniel Dick http://arxiv.org/abs/2606.02695v2 2026-06-05T12:42:21Z

Metamaterials can achieve exceptional functionality through careful engineering of their mesoscale structure. Although appropriately introduced irregularities can be advantageous, current approaches largely conform to regular structures to preserve tractability. Here, we contend that network theory, enriched with geometry and physics, provides a natural framework for designing metamaterials with controlled irregularities at relevant scales, thereby enabling the discovery of new property-enhancing structures. We examine how this augmented network theory can facilitate the creation of irregular metamaterials with enhanced or novel properties and how metamaterial research, in turn, is opening new directions in network science. Supported by machine learning and advanced self-assembly, the emerging field of irregular metamaterial networks is poised to transform inverse design and scalable manufacturing of novel materials.

2026-06-01T18:00:00Z 38 pages, 13 figures Thomas P. Wytock Chiara Daraio Heinrich M. Jaeger Christopher A. Schuh Lorenzo Valdevit Vincenzo Vitelli Adilson E. Motter http://arxiv.org/abs/2606.07197v1 2026-06-05T12:05:42Z

We study the low-energy excitations of the three dimensional spin glass through a large-scale Monte Carlo simulation on lattices up to $L=18$. We find smooth extrapolations down to zero temperature, which, in the case of the energy and of the link overlap, can be directly -- and favourably -- compared with previous investigations featuring ground states (i.e., at zero temperature). The best fit for the fractal dimension of the excitations is provided by Replica-Symmetry Breaking theory, but we also consider the alternative TNT description. The $P(q)$ is found to verify the Parisi-Toulouse temperature scaling. Our data provides a spectacular confirmation of the overlap-equivalence hypothesis.

2026-06-05T12:05:42Z 27 pages, 6 figures Claudio Chilin Enzo Marinari Víctor Martín-Mayor Giorgio Parisi Juan J. Ruiz-Lorenzo David Yllanes http://arxiv.org/abs/2605.22254v2 2026-06-05T07:25:17Z

We study a fully connected Hopfield-type associative memory network with online activity-dependent synaptic plasticity, where neural states and synaptic couplings coevolve during retrieval. Using the generating-functional formalism, we derive a dynamical mean-field theory (DMFT) in the large-system limit with extensively many stored random patterns, and show that the many-body dynamics reduces to an effective single-site stochastic process with colored Gaussian crosstalk noise and delayed feedback terms. Numerical solutions of the DMFT equations agree well with direct simulations. We find that moderate plasticity enlarges the basin of attraction and increases the maximum retrievable memory load by generating a positive delayed feedback that stabilizes retrieval against crosstalk noise. However, excessively strong plasticity causes the network to imprint the imperfect initial cue itself, leading to spurious attractors and degraded retrieval performance. Consequently, an optimal plasticity strength emerges from the trade-off between memory stabilization and premature cue imprinting. These results extend the DMFT description of associative memory to networks with coevolving neural and synaptic dynamics.

2026-05-21T09:59:44Z 37 pages, 10 figures Yoshinori Hara Yoshiyuki Kabashima http://arxiv.org/abs/2512.24869v2 2026-06-05T05:44:36Z

Topological markers provide an efficient and powerful characterization of topological features of many systems, especially when the translation symmetry is broken. Recently, a universal topological marker applicable in different symmetry classes of topological systems is proposed. However, it suffers from irregular behaviors at the boundary and its connection to other topological indexes remains elusive. In this work, we construct regularized universal topological markers that apply to Dirac systems by utilizing position operators that are compatible with periodic boundary conditions. The regularized markers eliminate the obstructive boundary irregularities successfully and give rise to the desired global topological invariants, such as the Chern number, consistently when integrated over all the lattice sites. Furthermore, the regularized form allows us to establish an explicit connection between the markers and some other known topological indices in two dimensions. For instance, it turns out to be equivalent to the Bott index in classes A, D, and C and equivalent to the spin Chern number in classes DIII and AII. We further examine the utility and stability of this marker in disordered scenarios. We find that its variance shows peaks at the phase boundaries, which promotes it as a useful indicator for detecting disorder-induced topological phase transitions.

2025-12-31T13:56:57Z 11 pages, 5 figures Phys. Rev. B 113, 214201 (2026) Yulin Qin Chang-An Li Jian Li 10.1103/yfxd-64tc http://arxiv.org/abs/2606.06925v1 2026-06-05T05:42:05Z

Computational models adequately representing phase transitions and evaluating the critical system parameters are essential for the understanding of the properties of a wide range of materials. Here we propose a machine learning (ML)-based approach to the identification of the critical point in anisotropic spin systems. Our approach implies training of a convolutional neural network (CNN) model from the correlation matrices obtained by Monte Carlo simulations. Next, the pretrained model is employed as a fast estimator of the critical temperature, which can be extracted in several complementary ways from the CNN model inference, this way improving the robustness of the analysis. The ML-based estimates obtained in this study are in very good agreement with the reference Monte Carlo simulation results, while computational costs are about 10x lower compared to the classical thermodynamic approach.

2026-06-05T05:42:05Z Alina A. Chubarova Ivan A. Mamonov Marina V. Mamonova Mikhail I. Bogachev Pavel V. Prudnikov