Power-law avalanches in RFIM survive only at low T or weak disorder
Temperature and disorder blur the size distribution in similar ways on the triangular lattice
· “Analysis of spin avalanches due to interplay of disorder and temperature”
cond-mat.dis-nn
Disordered Systems and Neural Networks
Glasses and spin glasses; properties of random, aperiodic and quasiperiodic systems; transport in disordered media; localization; phenomena mediated by defects and disorder; neural networks
sort pith recommended most recent
Simulations show graphene sheets forming in SiCN ceramics
ML potential enables 8000-atom simulations showing phase separation from amorphous matrix with network preserved and experimental match.
At n = d²/4, ellipsoid fitting flips from always to never
The exact threshold is set by the coordinates' fourth moment alone: heavier tails make fitting fail sooner.
A Krylov truncation is trustworthy for the round-trip transport time
Reliable window of a truncated quantum simulation ends at tau_m = sum 1/b_j, an on-the-fly stopping rule.
· “Krylov Break Times from an Inhomogeneous Lieb--Robinson Light Cone”
Below d^2/4, random points always fit an ellipsoid
Above that many points, almost surely no centered ellipsoid passes through them all.
· “The sharp SAT/UNSAT phase transition in random ellipsoid fitting”
Critical spin-glass overlaps settle on Airy_1 law
Two-replica overlap in Ising and spherical SK models gains a universal N^{-1/3} law; Talagrand's conjecture now follows.
· “Overlap distribution of the critical Sherrington-Kirkpatrick model”
Non-reciprocity ends aging, keeps FDT broken
In a spherical spin glass, any Γ<1 restores time-translational invariance, yet the steady state still violates fluctuation-dissipation.
· “Fluctuation-dissipation violations in mean-field non-reciprocal spin glasses”
Long-range couplings turn a log critical length into a power law
The ladder's critical length switches from log to algebraic to aspect-ratio scaling as couplings become nonlocal.
· “Nonlocality-induced critical-length hierarchy from non-Hermitian competition”
Anderson transition critical point shifts continuously with spectral dimension
In 3D fractal lattices, disorder strength at transition rises from 0 to 16.6 as spectral dimension increases from 2 to 3, with exponents set
· “Anderson Transition and Mobility Edges in a Family of 3D Fractal Lattices”
Two-layer nets cut critical slowing down to log scaling
Diffusion models for the O(n) model train in time that grows only logarithmically with size when locality is built into the architecture.
Local quantum information advances ballistically
Scale-resolved information profiles in Haar and Clifford circuits collapse onto a shifted Tracy-Widom curve.
· “Dynamics of local quantum information in random unitary circuits”
A looped transformer that repeatedly applies the same network can solve problems using…
Training selects one of several dynamical phases in looped transformers, and only the fold phase obeys the parameter-free compute-scaling…
· “Dynamical phase selection controls compute scaling in looped transformers”
Weevil blue comes from ring and pore resonances
Matching simulated spectra identifies short-range ring and pore disorder as the control, with long-range order playing a minor role.
Reentrant p-spin glass boundary forces a zero-overlap temperature pair
The boundary's curvature equals a specific-heat difference, and it is positive from p=3 to p=12, so cooling erases ferromagnetic order.
· “Reentrance and temperature chaos in the p-spin Ising spin glass”
Dissipation and memory time bound the error of computation
Accurate computation in changing environments needs dissipation or long-lived memory.
· “Entropy Production Bounds the Accuracy of Computation in Markov Networks”
Dissipation alone creates quantum chaos in a cavity fermion gas
Spontaneous emission or multimode loss pushes ultracold fermions to random-matrix chaos; rank is the dial.
· “Dissipation-induced Sachdev-Ye-Kitaev physics in many-body cavity quantum electrodynamics”
Inertia makes high-dimensional sync transitions jump abruptly
Kuramoto model with inertia shows hysteresis beyond a dimension-dependent threshold; odd dimensions still jump at zero inertia.
· “Inertial synchronization of networked oscillators in arbitrary dimensions”
Random bonds can't knock Lévy-SAW out of its universality class
Monte Carlo and field theory show the graph's random bonds leave the clean long-range exponents intact.
Strong transverse fields erase replica order in quantum glasses
A universal Hessian bound forces the quantum replica overlap to vanish at strong transverse field.
· “Correlation inequalities for transversal field models with application to quantum glasses”
Two-timescale memory reaches target loss 55 steps sooner than Muon's record
Swapping only the momentum kernel for fast and slow branches accelerates a frozen LLM training benchmark.
· “A Physical Response-and-Memory Model for Muon Optimization”
The paper simulates Q-learning agents playing prisoner's dilemma on networks and varies…
Q-learning agents in spatial prisoner's dilemma cooperate most when they observe roughly three to four neighbors, a non-monotonic effect…
· “Evolution of cooperation with Q-learning: how much information do we need?”
Loss landscape contracts near solutions as data complexity rises
Higher label mixing shrinks the volume of low-loss parameters close to the trained solution, not uniformly far away.
· “Dataset Complexity Shapes Finite-Distance Loss Geometry in Neural Networks”
Flux-tuned mobility edges are exact in a quasiperiodic chain
Adding an auxiliary lattice with synthetic flux makes localization boundaries in these chains exactly solvable and continuously tunable.
· “Engineering exact mobility edges in quasiperiodic Aharonov-Bohm chains”
Periodic driving can't create a new entanglement phase in 1D fermions
The drive only rescales one conductivity parameter, so monitored chains still saturate to area-law entanglement for every measurement rate.
Compress wide networks to a width set by data dimension
Hidden layers shrink to a width fixed by input dimension and the error budget, with no fine-tuning.
· “Width-Independent Compressibility of Deep Neural Networks”
Short-chain-trained force field reaches 257-unit polyethylene
Deep-potential single-chain statistics match good-solvent theory, with scaling exponent v=0.61.
Finite-width neural networks must drop a QFT axiom
Exact QFT correlation functions force infinite width; finite networks lose reflection positivity or clustering.
One gain unit makes scale-free brain dynamics generic
A single normalization-like feedback loop pins the top eigenvalue to the stability edge for any coupling strength.
· “A minimal mechanism to generate long timescales without fine tuning”
100 snapshots pinpoint the ergodicity-breaking transition
Snapshot dimension and scale-free networks separate ergodic, critical, and non-ergodic regimes of the quantum sun model.
· “Many-body ergodicity breaking from wavefunction snapshots”
One exact identity splits MCMC relaxation into cooling plus entropy
The accept/reject bit acts as a demon, and the balance shows exactly which part of its information speeds up sampling.
· “Maxwell's Demon in Markov Chain Monte Carlo: Cooling Information Flow and Entropy Balance”
Dynamic correlations peak, then shrink, at the glass crossover
In the beta-relaxation window, the correlation length peaks near Tc and falls on further cooling, even as local fluctuations intensify.
· “Non-Monotonic Dynamical Correlations Across The Glass Crossover”
Gradient descent hits algorithmic energy floor in one glass phase
Exact asymptotic energies classify one-step, two-step, and continuous aging as the interaction mix changes.
· “Aging Phase Diagram and Exact Asymptotic Energies of Mixed Spherical Spin Glasses”
Potts learning hits a tricritical point with exponent 4/15
There, the Edwards-Anderson correlator decays with the exact Potts spin exponent, fixing the universality class.
· “Learning Potts Models and Z₃ Toric Codes: Higher and Ordinary Nishimori Criticality”
Quantum learning peaks before chaos
One coupling angle sweeps from weak to scrambled; the learning sweet spot sits in between, and scales with size.
RL turns Trotter approximation errors into self-correction
An 8-spin policy controls 100-spin evolution from energy and variance alone, beating fixed-step and greedy Trotter.
· “Reinforcement Learning to Harness Approximation Errors for Long-Time Quantum Simulation”
Reputation alone flips the prisoner's dilemma to full cooperation
A trust-weighted blend of own and neighbors' payoffs creates a discontinuous, bistable cooperative phase in a spatial Q-learning model.
· “Emergence of cooperation: A reputation-modulated reinforcement learning”
Dynamic pseudogap spectra are a sum of Bogoliubov histories
A parameter-free decoherence scale from the field correlations picks resolved sidebands or a static-like gap.
Random neural networks are lattice gauge theories
With layerwise symmetry gauged, weights are gauge fields and all loop corrections to the neuron propagator are resummed.
Two entropy rules build working proteins
A layer-restricted Boltzmann machine constructs functional chorismate mutases up to 53% divergent from nature.
A single damping constant tames RNN quantum-state training
The fix lets minSR natural-gradient optimization beat Adam on 1D spin models using only ~100 samples.
· “Quantum Geometric Tensor Preconditioning for Stable Training of Recurrent Neural Quantum States”
Simplicial closure fragments explosive cooperation transitions
Closure turns one first-order-like cooperation transition into a spread of metastable states.
Dephasing creates mobility edges in a quasicrystal
In a dimerized off-diagonal quasiperiodic chain, pure dephasing splits the relaxation spectrum into extended, critical, and localized modes.
· “Dephasing-induced distinct mobility edges in a dimerized off-diagonal quasicrystal”
Neural screen finds hard-carbon sodium anodes above 800 mAh/g
A surrogate trained on 64 structures ranks 13,000+ and points to low-density, porous carbons as the sweet spot.
Singular kicked rotor reproduces PRBM Anderson transition
A power-law or logarithmic kick makes a single Floquet map show localized, critical, and extended phases without any disorder.
· “Long-range Nonlinear Sigma Model for a Singular Quantum Kicked Rotor”
Chaotic edge states resist perturbations with D^(1/3) rigidity
Lowest-lying random-matrix eigenstates are shown to be D^(1/3)-robust, with symmetry-dependent universal distributions.
Scarred spin chain doubles its period without disorder
Quantum scars keep the oscillation alive for a size-dependent lifetime, then thermalization melts it.
· “Scarred discrete time crystal in a periodically driven dimerized spin chain”
Billion-node composites solved in about an hour by FNO swarm
Unit-cell neural operators plus Schwarz iteration match nonlinear FEM accuracy at a fraction of the cost.
Neural blocks upscale composites from micro to macro scale
A graph-assembled network self-trains larger blocks, reaching a 1000x1000 mosaic at ~3% error versus direct FEM.
· “Neural-Embedded Graphical Model for Self-Consistent Hierarchical Upscaling of Complex Composites”
The 1/Γ resonance law ends in a first-passage front
Finite length cuts narrowest widths at a computable scale; same process yields Wigner delay and boundary force.
The multifractal phase, not the metallic, combines sliding and spreading
Nonreciprocal hopping in a hexagonal Harper chain makes the multifractal regime superdiffusive in both measures.
· “Localization and Transport in a Non-Hermitian Hexagonal Harper Model”
Demographic noise tames chaotic ecosystems by pruning rare species
A rule-based model shows survivors satisfy May's stability criterion, then fade as a power law.
· “Extinction drives emergent metastability in complex ecosystems”
Nonlinear wave network measures its own gradient in one extra pass
A phase-engineered probe lets the hardware measure all sensitivities at once, no digital twin needed.
· “In-situ Adjoint Wave Control in Reconfigurable Non-Hermitian Nonlinear Systems”
One symmetry unifies MLPs, attention, and convolutions
A single quadratic model predicts plateaus and power-law scaling from the architecture's structure matrix.
· “Neural Quadratic Forms: A Unified Minimal Model for Sudden Learning and Scaling Laws”
Magnetic anisotropy sets Néel temperature in diluted antiperovskites
In six 1/3-filled compounds, the Néel temperature climbs from 29 K (Mn) to 50 K (Fe) to 92 K (Co).
Graph metrics can pick better laser vision networks 3,000 times faster
Two connectivity measures predict lasing physics and image-classification accuracy, letting evolution skip costly simulation.
· “Graph-theoretic design of lasing networks for physical vision”
Gap grows as superconductivity dies in disordered films
Tunneling and simulations show the gap persists and widens in the insulator, evidence for localized Cooper pairs.
· “The nature of the "pseudogap" in the insulating phase of highly disordered superconductors”
One mode pair decides when non-Hermitian spectra turn complex
Threshold equals the ratio of reference detuning to projected gain-loss coupling, solved without expansion
· “Exceptional activated mode theory for generalized real-complex transitions”
D-wave superconductivity wins in the doped Hubbard model
A translation-symmetric wave function beats stripe states and extrapolates to a finite pairing order.
· “Superconductivity in the t-t' Hubbard Model from Symmetry-Preserving Neural-Network Quantum States”
Zero-crossing hopping makes all eigenstates share one envelope
The envelope is log-correlated: high stiffness localizes, low stiffness gives a multifractal.
· “Logarithmically Correlated Landscapes and Localization in Non-Hermitian Quasicrystals”
Physical devices can now compute their own exact gradients
Reciprocity lets linear hardware train in one extra run; nonlinear trajectories need a time-reversal mirror.
Ring-averaged normalization fixes nanoscale electron correlation artifacts
Time- and angle-averaged intensity removes false correlations and recovers true relaxation times.
· “Accurate Evaluation of Nanoscale Spatiotemporal Dynamics with Electron Correlation Microscopy”
A spectral drop reveals hidden stealthy disorder parameters
Transport data alone recovers the hidden stealthiness parameter χ and disorder strength W.
· “Inferring stealthy hyperuniform correlations from quantum transport”
Skin localization persists beyond point-gap topology
Tuning one coupling parameter switches boundary pile-up into Anderson localization and back again
· “Skin-Anderson Localization Transition in Strongly Coupled Disordered Non-Hermitian Chains”
Rare strong hops make a single particle mimic many-body subdiffusion
A single-particle model shows a subdiffusive window that widens near localization, benchmarking many-body studies.
· “Analytical Theory for Anomalous Diffusion in the Anderson Model with Heavy Tails”
Sliding-window decoding obeys reaction-diffusion kinetics
A single scaling curve links real-time and static decoding: the window size W sets how long logical memory survives.
Delays make heterogeneous directed networks the best synchronizers
A generalized master stability function shows that hub-like master-slave wiring beats all-to-all in delayed systems.
· “Generalized Master Stability of Heterogeneous Delay-Coupled Networks”
Attention condenses past a finite sharpness threshold
With token count and dimension growing together, clusters appear at beta = O(1), not at the sqrt(log N) static onset.
· “Clustered Attractor Manifolds and Dynamical Condensation in Self-Attention”
Ring-width diversity cuts error in magnetic reservoir computer
Multiple readout channels from graded nanorings expand the usable state space, not just the parameter count.
· “Reservoir Computing with Heterogeneous Magnetic Metamaterials”
Pairs stay extended while singles localize
Opposite leg potentials cancel for same-rung pairs, making internal configuration a reversible localization switch.
· “Composite-State Localization Beyond the External Landscape in Non-Hermitian Quasicrystals”
Flat Chern bands localize by their quantum geometry
A transfer-matrix study ties localization length to band geometry and predicts a quantum-geometric mobility edge.
· “Quantum geometric localization length and localization criticality in an ideally flat Chern band”
One-way chat drives an AI into alien behavior absent in isolation
Listening, not identity, decides which identical AI changes; who listens governs who leaves its baseline.
· “Interaction Creates Dynamical AI Behavior Absent in Isolation”
Volumetric strain, not shear, sets vacancy formation cost
Neural network matches atomistic vacancy energies to 0.026 eV and shows local chemistry controls the spread.
Imperfect generative models can still yield exact Monte Carlo samples
A tutorial review shows how flows and diffusion models can sample multimodal targets without bias.
· “Leveraging generative models to assist Monte Carlo sampling”
Correlated disorder preserves many-body localization in qubit chains
Flux-induced coupling noise barely shifts the MBL critical point, easing design of localized transmon processors.
· “Many-Body Localization Induced by Correlated Disorder in Interacting Superconducting Qubits”
D-shaped mobility edge found in open many-body spin chain
One phase boundary separates extended states, many-body localization, and the non-Hermitian skin effect in a single model.
· “Many-Body Mobility Edge and Non-Hermitian Skin Effect in an Interacting Quasi-Periodic Spin Chain”
Deep linear nets learn features one phase transition at a time
Each data direction turns on at its own critical regularization strength, with known critical exponents.
Spectral width below 1 makes Ising free energy a local graph property
Graphs with matching local structure then share one free energy density, even when couplings are random or antiferromagnetic.
Local correlators wander with t^{2/3}
In noisy random circuits the center of a one-particle correlator wanders on a t^{2/3} scale, not diffusively.
· “KPZ Superdiffusion of Local Correlators in Diffusive Random Quantum Circuits”
Neural flux attachment learns condensates and chiral edge states
A fixed pairwise vortex converts a fermionic transformer into exact bosons, reaching Kalmeyer-Laughlin edges at N=20.
· “Neural Flux Attachment: From Bose Condensates to Chiral Topological Matter”
Polynomial decoders entangle exponentially with tiny latent width
For n Bell pairs, a degree-n decoder needs just 0.29n latent variables instead of 2^n Schmidt coefficients.
