1572 papers
Statistical mechanics explores how the chaotic motion of countless tiny particles gives rise to the predictable laws governing heat, pressure, and phase transitions. This field bridges the gap between the microscopic world of atoms and the macroscopic reality we experience daily, offering deep insights into why materials behave the way they do.
On Gist.Science, we process every new preprint in this category as it appears on arXiv to make these complex findings accessible to everyone. For each paper, we provide both a plain-language explanation for the curious reader and a detailed technical summary for specialists, ensuring that groundbreaking research is never lost behind a wall of jargon.
Below are the latest papers in statistical mechanics, freshly curated and summarized to help you understand the cutting edge of this fascinating discipline.
This paper provides a complete mechanistic characterization of in-context learning in transformers trained on discrete Markov chains, revealing four algorithmic phases implemented by distinct multi-layer subcircuits and identifying the specific data diversity thresholds and training dynamics that govern the transition between memorization and generalization strategies.
This paper theoretically investigates steady-state and non-equilibrium dynamics in noisy networks by deriving conditions for equilibrium, analyzing non-equilibrium steady states through probability currents and drift velocities, establishing a general fluctuation-dissipation relation, and demonstrating that conventional overdamped Brownian dynamics is a special case of this broader framework.
This paper presents an exact information-geometric decomposition of unsupervised learning generalization error into model error, data bias, and variance, applying the framework to -PCA to derive an optimal rank selection rule and a three-regime phase diagram that balances model complexity against data noise.
This paper demonstrates that extending the Biswas-Chatterjee-Sen (BChS) model to include group interactions of size shifts the critical noise threshold monotonically toward as increases, yet the system retains the mean-field Ising universality class with unchanged critical exponents for all interaction sizes.
This paper demonstrates that generalized symmetries, including higher-form, subsystem, gauge, and non-invertible symmetries, can induce exponential Hilbert space fragmentation, thereby challenging the assumption that such fragmentation necessarily implies ergodicity breaking and revealing its role in disorder-free localization.
This paper classifies the computational complexity of 2-local Hamiltonian problems with positive symmetric interactions into three distinct phases—QMA-complete, StoqMA-complete, and a new EPR* class—by utilizing perturbative gadgets and renormalization-group-like flows to demonstrate that EPR* likely represents the transition point between easy and hard local Hamiltonians.
This paper investigates normal mode analysis in relativistic massive transport by demonstrating how particle mass induces coupling between sound and heat channels, alters the branch cut structure responsible for Landau damping, and leads to non-monotonic dependencies in the critical wavenumbers for collective modes.
This paper proposes two bio-inspired learning algorithms for one-dimensional time series based on the statistical-mechanical properties of the Loewner equation—specifically Gaussian process regression via driving force normality and a fluctuation-dissipation relation for sensitivity analysis—which are numerically validated on neuronal dynamics and discussed in the context of biological self-organization.
This paper investigates the dynamics of nonstabilizerness (magic) in monitored quantum circuits, revealing that computational basis measurements exponentially suppress magic through Clifford scrambling, whereas rotated non-Clifford measurements can both generate and sustain magic, leading to distinct relaxation behaviors and diagnostic sensitivities between stabilizer nullity and Stabilizer Rényi Entropies.
This paper demonstrates that in fractonic systems where the -th multipole moment of the order parameter is conserved, domain coarsening after a quench follows an anomalously slow growth law of , thereby establishing a new family of non-equilibrium universality classes characterized by a cascade of dynamical critical exponents.