https://arxiv.org/api/2go3BkeNiaslD/3bmWzooVqhta8 2026-06-10T10:46:00Z 26559 120 15 http://arxiv.org/abs/2507.10892v2 2026-05-27T14:31:36Z

Understanding how quantum chaotic systems generate entanglement can provide insight into their microscopic chaotic dynamics and can help distinguish between different classes of chaotic behavior. Using von Neumann entanglement entropy, we study a nonentangled state evolved under three variants of the Sachdev-Ye-Kitaev (SYK) model with a finite number of Majorana fermions $N$. All the variants exhibit linear entanglement growth at early times, which at late times saturates to a universal value consistent with random matrix theory (RMT), but their growth rates differ. We interpret this as a large-$N$ effect, arising from the enhanced non-locality of fermionic operators in SYK and binary SYK, absent in spin operators of the spin-SYK model. Numerically, we find that these differences emerge gradually with increasing $N$. Although all variants are quantum chaotic, their entanglement dynamics reflect varying scrambling rates and indicate that the entanglement production rate serves as a fine-grained probe of scrambling beyond conventional measures. To probe its effect on thermalization properties of these models, we study the two-point autocorrelation function, finding no differences between the SYK variants, but deviations from RMT predictions for $N \geq 24$, particularly near the crossover from exponential decay to saturation regime.

2025-07-15T01:13:31Z Published version Phys. Rev. E 113, L052204 (2026) Tanay Pathak Masaki Tezuka 10.1103/9jbx-9kln http://arxiv.org/abs/2605.00435v2 2026-05-27T13:56:08Z

Mode collapse is a persistent challenge in generative modeling and appears in autoregressive text generation as behaviors ranging from explicit looping to gradual loss of diversity and premature trajectory convergence. We take a dynamical-systems view and reinterpret mode collapse as reduced state-space accessibility caused by *geometric collapse*: during generation, the model's internal trajectory becomes confined to a low-dimensional region of its representation space. This implies mode collapse is not purely a token-level phenomenon and cannot be reliably solved by symbolic constraints or probability-only decoding heuristics. Guided by this perspective, we propose *Reinforced Mode Regulation* (RMR), a lightweight, online state-space intervention that regulates dominant self-reinforcing directions in the Transformer value cache (implemented as low-rank damping). Across multiple large language models, RMR substantially reduces mode collapse and enables stable generation at extremely low entropy rates (down to 0.8 nats/step), whereas standard decoding typically collapses near 2.0 nats/step.

2026-05-01T06:12:05Z Accepted to ICML 2026 Xin Du Kumiko Tanaka-Ishii http://arxiv.org/abs/2605.28395v1 2026-05-27T12:32:20Z

We present a systematic benchmark of MACE potentials for iron-nickel alloys, focusing on structural, elastic, magnetic, and finite-temperature properties relevant to phase stability. The reference dataset comprises spin-polarized PBE density functional theory (DFT) calculations for chemically disordered special quasirandom structures (SQS), spanning compositions, bcc and fcc crystal structures, and volumetric and shear deformations. A system-specific MACE-sqs model trained on this dataset achieves validation errors of 2.0 meV/atom for energies and 24.3 meV/Angstrom for forces. Compared with several MACE foundation models, including models trained with Hubbard U corrections, MACE-sqs gives the most consistent agreement with DFT and experiment for equations of state, equilibrium volumes, elastic constants, and thermal expansion trends in bcc and fcc Fe-Ni alloys. For the bcc-to-hcp transition, MACE-sqs predicts a pure-Fe transition pressure closer to experiment than the tested foundation models, but all models predict an incorrect increase of transition pressure with Ni content. This failure indicates that high-pressure magnetic collapse and composition-dependent magnetoelastic effects are not yet fully captured. Overall, targeted SQS-based training substantially improves the accuracy of MACE potentials for Fe-Ni alloys, while phase stability under magnetic collapse remains a key limitation for future model development.

2026-05-27T12:32:20Z Kushal Ramakrishna Mani Lokamani Attila Cangi http://arxiv.org/abs/2605.11637v2 2026-05-27T04:52:50Z

X-ray computed tomography (CT) reveals the materials' internal structures non-destructively from a tilt series of projected images. Filtered back projection (FBP) is a widely-adopted reconstruction algorithm in CT owing to its small computational cost. Under low-dose or sparse-view conditions, however, FBP often amplifies noise, severely degrading the reconstructed images. In this study, we evaluated the performance of a Bayesian CT reconstruction algorithm based on the Markov random field model under such adverse conditions. Through simulations, we demonstrated that the proposed algorithm shows higher reconstruction performance than FBP under both low-dose and sparse-view conditions. The hyperparameters are estimated by minimizing the Bayesian free energy, enabling adaptive reconstruction that reflects the noise characteristics of the observed projection data. These results suggest that the proposed algorithm can broaden the applicability of CT to dose-sensitive applications and time-constrained measurements, where only limited observed projection data are available.

2026-05-12T06:58:44Z 18 pages, 7 figures Taiga Shimomiya Taichi Kusumi Masayuki Uesugi Akihisa Takeuchi Yuki Sada Hayaru Shouno Masato Okada http://arxiv.org/abs/2512.18783v2 2026-05-27T01:10:57Z

Noisy monitored quantum circuits have emerged as a versatile and unifying framework connecting quantum many-body physics, quantum information, and quantum computation. In this review, we provide a comprehensive overview of recent advances in understanding the dynamics of such circuits, with an emphasis on their entanglement structure, information-protection capabilities, and noise-induced phase transitions. A central theme is the mapping to classical statistical models, which reveals how quantum noise reshapes dominant spin configurations. This framework elucidates universal scaling behaviors, including the characteristic $q^{-1/3}$ entanglement scaling with noise probability $q$ and distinct timescales for information protection. We further highlight a broad range of constructions and applications inspired by noisy monitored circuits, spanning variational quantum algorithms, classical simulation methods, mixed-state phases of matter, and emerging approaches to quantum error mitigation and quantum error correction. These developments collectively establish noisy monitored circuits as a powerful platform for probing and controlling quantum dynamics in realistic, decohering environments.

2025-12-21T15:55:12Z 20 pages, 17 figures J. Phys.: Condens. Matter 38 203001 (2026) Shuo Liu Shao-Kai Jian Shi-Xin Zhang 10.1088/1361-648X/ae6a58 http://arxiv.org/abs/2602.02855v2 2026-05-26T21:26:39Z

Pre-training on a source task is usually expected to facilitate fine-tuning on similar downstream problems. In this work, we mathematically show that this naive intuition is not always true: excessive pre-training can computationally slow down fine-tuning optimization. We study this phenomenon for low-rank adaptation (LoRA) fine-tuning on single-index models trained under one-pass SGD. Leveraging a summary statistics description of the fine-tuning dynamics, we precisely characterize how the convergence rate depends on the initial fine-tuning alignment and the degree of non-linearity of the target task. The key take away is that even when the pre-training and downstream tasks are well aligned, strong pre-training can induce a prolonged search phase and hinder convergence. Our theory thus provides a unified picture of how pre-training strength and task difficulty jointly shape the dynamics and limitations of LoRA fine-tuning in a nontrivial tractable model. On the practical side, we empirically show that our theoretical findings extend beyond our toy model and remain relevant in the context of a vision-transformer model trained on real data.

2026-02-02T22:02:52Z 38 pages, 14 figures Gibbs Nwemadji Bruno Loureiro Jean Barbier http://arxiv.org/abs/2506.04219v2 2026-05-26T21:10:01Z

Multicellular organisms contain a wide variety of highly specialized cell types. The consistency and robustness of developmental trajectories suggest that complex gene regulatory networks effectively act as low-dimensional cell fate landscapes. Prior work inspired by dynamical systems theory argues that cell fate transitions fall into universal decision-making classes, but the theory connecting these geometric landscapes to high-dimensional gene expression space is still in its infancy. Here, we introduce a phenomenological model that identifies experimental signatures of decision-making classes in single-cell RNA-sequencing time-series data. The model combines low-dimensional gradient-like dynamics with high-dimensional Hopfield networks to capture the interplay between cell fate, gene expression, and signaling. We apply the framework to experimental mouse data on maturing lung alveolar cells and lineage-traced hematopoietic differentiation and show that the measured cell fate dynamics are consistent with developmental landscapes containing intermediate progenitors and saddle points. We further show that the framework can be used to understand spatial patterning and cell fate organization, focusing on Notch signaling in lung airways. Together, these results provide evidence that collective transcriptomic dynamics carry signatures of landscape features associated with universal decision-making classes.

2025-06-04T17:57:54Z 19 pages, 4 figures Maria Yampolskaya Laertis Ikonomou Pankaj Mehta http://arxiv.org/abs/2605.27524v1 2026-05-26T18:01:01Z

We examine the emergence of topological Anderson insulating phases in the spinful Haldane model with Hubbard and next-neighbor density-density interactions, subject to Anderson disorder. Using finite-size exact diagonalization, we characterize the phases that arise from the interplay between topology, interactions, and disorder. In addition to standard $C=2$ topological Anderson phases, we observe an antiferromagnetic $C=1$ topological Anderson phase, consistent with the antiferromagnetic quantum anomalous Hall insulator previously identified in the clean model at finite staggered mass. We further analyze these phases using a neural network trained on the exact diagonalization data. Our results support the hypothesis that an explicit charge imbalance is required to induce the $C=1$ phase, generated by Anderson disorder rather than by a staggered mass.

2026-05-26T18:01:01Z Alejandro J. Uría-Álvarez Roser Valentí http://arxiv.org/abs/2605.27514v1 2026-05-26T18:00:04Z

The configuration space of causal sets is vast. It is a critical goal to map out this space. Here, we take a practical step towards this goal. We investigate nine classes of causal sets, most of them not studied before. These include manifoldlike causal sets with inhomogeneous Ricci curvature, both topologically trivial and nontrivial. We also study classes of non-manifoldlike causal sets, including lattices, layered orders as well as Lorentzian quasicrystals. Finally, we study classes of causal sets that are not manifoldlike, but are expected to become manifoldlike under a suitable coarse-graining process. We use this broad range of distinct classes of causal sets as a testbed for observables. Rather than focusing on continuum-geometry inspired observables, such as curvature invariants, which often exhibit large fluctuations and are computationally very expensive, we focus on graph observables, including some observables that constitute subgraph statistics and some that are global. We find that three observables, namely the link degree distribution, the eigenvalues of the graph Laplacian of the symmetrized Hasse diagram and the recently proposed abundance of causal intervals, can distinguish between the distinct classes of causal sets. This is made possible by the small fluctuations that these observables have in most classes.

2026-05-26T18:00:04Z 48+14 pages, 21 figures Astrid Eichhorn Harald Mack Kim Tuyen Le Fabian Wagner http://arxiv.org/abs/2511.12647v3 2026-05-26T17:49:22Z

We investigate quantum inverse problems in one-dimensional (1D) electronic disordered systems strongly coupled to optical cavities. More specifically, we consider the Anderson and the Aubry-Andre-Harper models connected to electronic reservoirs and embedded in a single-mode optical cavity. The light-matter interaction enables photon-assisted hopping processes that significantly modify the transmittance spectrum. Within the nonequilibrium Green's function formalism, we implement an inversion-based approach capable of accurately extracting the electron-photon coupling strength directly from transmittance spectra. While cavity coupling acts as a minor perturbation within the Anderson model, yielding broad yet precise parameter estimates, its influence is markedly different in the Aubry-André-Harper model. The latter exhibits a sharp metal-insulator transition in 1D, thus resulting in more pronounced cavity-induced spectral changes. This renders even more accurate inverse solutions, offering unparalleled precision in the characterization of low-dimensional disordered systems. Altogether, our results demonstrate that the quantum inverse problem provides a robust diagnostic tool for quantum materials, particularly effective for systems exhibiting metal-insulator transitions.

2025-11-16T15:29:06Z 13 pages, 9 figures Thales F. Macedo Julián Faúndez Antônio S. Coelho Caio Lewenkopf Mauro S. Ferreira Felipe A. Pinheiro Natanael C. Costa http://arxiv.org/abs/2502.19063v5 2026-05-26T17:39:39Z

We show that a simple nonlinear differential equation (originally studied in the physics of disordered systems) is able to mathematically describe the global population growth over the past 12000 years. Different regimes of population growth since the early Neolithic until today are shown to be all solutions to the same nonlinear differential equation in its various limits. These also include the well-known Malthus (exponential) and Verhulst (logistic) growth regimes, as well as von Foerster's ``doomsday'' formula. All these limits correspond to neglecting higher-order terms in a more general nonlinear dynamic model described by the proposed nonlinear differential equation. While the older models may provide valid fittings to limited time intervals in the global population growth curve in time, their clearly approximate nature prevents them from being predictive over longer periods of time. The proposed comprehensive solution of the proposed model is instead well suited to provide predictions for future scenarios. These include a scenario where the global population could halve as early as 2064 under a deliberately conservative, worst-case assumption that carrying-capacity constraints become abruptly active today.

2025-02-26T11:41:45Z Chaos, Solitons & Fractals 209, 118542 (2026) Alessio Zaccone Kostya Trachenko 10.1016/j.chaos.2026.118542 http://arxiv.org/abs/2605.20358v2 2026-05-26T14:04:32Z

Polymer-derived ceramics combine the thermal stability of ceramics with the versatile properties of carbon domains, but modeling their atomic-scale evolution during processing remains elusive due to the limitations of traditional computational methods. To address this issue, here we develop and apply a machine learning interatomic potential for silicon carbonitride-based (Si-C-N-H) systems, trained on a diversified database of over 9000 configurations -- including amorphous models, high-temperature states, surfaces, and crystal structure predictions -- to capture the full complexity of these materials. This potential enables large-scale molecular dynamics simulations of 8000-atom systems revealing the atomic-scale evolution of the polymer-derived ceramic during thermal treatment. A key result of this work is the occurrence of a phase separation where carbon domains progressively nucleate from the amorphous SiCN matrix during thermal processing, forming distinct graphene-like sheets while preserving the integrity of the ceramic network. The resulting models reproduce the experimental atomic pair distribution functions with exceptional fidelity, validating our approach and providing microscopic explanations for the material unique combination of ceramic and graphitic properties. In this process, defective 5- and/or 7-member carbon rings, mediate the transformation to stable 6-member aromatic structures. These findings offer new atomic-scale insights into the thermal stability and structural transformation pathways of polymer-derived ceramics, while our methodology opens avenues for studying complex amorphous systems with first-principles accuracy at experimentally relevant scales.

2026-05-19T18:12:25Z Fabien Mortier Sylvian Cadars Olivier Masson Mauro Boero Guido Ori Yun Wang Samuel Bernard Assil Bouzid http://arxiv.org/abs/2605.27006v1 2026-05-26T13:26:36Z

Sampling from learned high-dimensional distributions is a foundational computational problem. We introduce U-turn chains: Markov chains obtained by iterating short forward-backward steps of a diffusion model, in which each step proposes a move that remains on the learned data manifold and, paired with a Metropolis-Hastings correction, samples from energy-modified targets. For synthetic languages, we show that minimal U-turn dynamics undergoes an ergodicity-breaking phase transition driven by fragmentation of the data manifold; ergodicity is restored at larger U-turn magnitude. In the non-ergodic regime, low-level features relax faster than high-level ones, an ordering that inverts only at sufficiently large U-turn magnitude. We test these predictions on natural language and natural images. In both modalities, minimal U-turns relax slowly, especially for high-level features approximated by deep representations in CNNs or LLMs. The layer-ordering inversion appears only at large noise when mixing is efficient -- signatures consistent with strongly constrained, weakly mixing local dynamics. We discuss the implications of these results for sampling with diffusion models.

2026-05-26T13:26:36Z Hyunmo Kang Noam Itzhak Levi Corinna Elena Wegner Daniel J. Korchinski Matthieu Wyart http://arxiv.org/abs/2605.26973v1 2026-05-26T12:58:48Z

Neural networks are known to develop latent representations that are $aligned$, namely structurally similar across networks trained with different architectures, training protocols, or training datasets. We study this phenomenon in a controlled setting, where we train an ensemble of networks on regression and classification tasks using training sets perturbed by independent realizations of a noise process. We show that the signal-to-noise ratio (SNR) and the training sample size influence the alignment in qualitatively similar ways in networks trained on real-world datasets and in an extremely simple $linear$ network with a single hidden layer, for which the alignment can be estimated analytically. Across linear and nonlinear networks, regression and classification tasks, and both synthetic and real-world data, we consistently observe that alignment varies monotonically with SNR but non-monotonically with training sample size. In particular, the alignment is minimized near the interpolation threshold, and a stronger alignment does not necessarily correspond to better generalization error. These findings reveal a non-trivial dependence of alignment on data quality and quantity, decoupled from generalization performance.

2026-05-26T12:58:48Z Ali Hussaini Umar Alessandro Laio http://arxiv.org/abs/2605.26767v1 2026-05-26T09:38:03Z

We performed parallel measurements of heat effects and shear modulus relaxation for glassy Te$_{75}$Ge$_{15}$Ga$_{10}$ taken as a representative of practically important non-metallic glasses with covalent bonding. It is shown that the heat effects occurring upon heating are quantitatively linked to the shear moduli in the glassy and crystalline states and their temperature derivatives as implied by Eq.(1), which was originally derived for metallic glasses. This relationship provides a good description of exo- and endothermal reactions using shear modulus relaxation data as an input. This is the first application of this approach to a non-metallic glass with directional interatomic bonding. The obtained results suggest that relaxation phenomena are governed by elastic dipoles -- atomic configurations with the symmetry lower than that of surrounding matrix.

2026-05-26T09:38:03Z 12 pages, 3 figures G. V. Afonin R. A. Konchakov D. A. Chareev S. A. Badmaeva R. S. Khmyrov A. N. Vasiliev N. P. Kobelev V. A. Khonik