arXiv subjects

Federico Ghimenti

Publications and source records attributed to Federico Ghimenti.

14 recordsLinked to original sources

arXiv2026

Contrastive Concept-Tree Search for LLM-Assisted Algorithm Discovery

Large language Model (LLM)-assisted algorithm discovery is an iterative, black-box optimization process over programs to approximatively solve a target task, where an LLM proposes candidate programs and an external evaluator provides task feedback. Despite intense recent research on the topic and promising results, how can the LLM internal representation of the space of possible programs be maximally exploited to improve performance is an open question. Here, we introduce Contrastive Concept-Tree Search (CCTS), which extracts a hierarchical concept representation from the generated programs and learns a contrastive concept model that guides parent selection. By reweighting parents using a likelihood-ratio score between high- and low-performing solutions, CCTS biases search toward useful concept combinations and away from misleading ones, providing guidance through an explicit concept hierarchy rather than the algorithm lineage constructed by the LLM. We show that CCTS improves search efficiency over fitness-based baselines and produces interpretable, task-specific concept trees across a benchmark of open Erd\H{o}s-type combinatorics problems. Our analysis indicates that the gains are driven largely by learning which concepts to avoid. We further validate these findings in a controlled synthetic algorithm-discovery environment, which reproduces qualitatively the search dynamics observed with the LLMs.

cs.LG↗

arXiv2025

Short-term plasticity recalls forgotten memories through a trampoline mechanism

We analyze continuous Hopfield associative memories augmented by additional, rapid short-term associative synaptic plasticity. Through the cavity method, we determine the boundary between the retrieval and forgetting, or spin-glass phase, of the network as a function of the fraction of stored memories and the neuronal gain. We find that short-term synaptic plasticity yields marginal improvements in critical memory capacity. However, through dynamical mean field theory, backed by extensive numerical simulations, we find that short-term synaptic plasticity has a dramatic impact on memory retrieval above the critical capacity. When short-term synaptic plasticity is turned on, the combined neuronal and synaptic dynamics descends a high-dimensional energy landscape over both neurons and synapses. The energy landscape over neurons alone is thus dynamic, and is lowered in the vicinity of recent neuronal patterns visited by the network, just like the surface of a trampoline is lowered in the vicinity of regions recently visited by a heavy ball. This trampoline-like reactivity of the neuronal energy landscape to short-term plasticity in synapses can lead to the recall of stored memories that would otherwise have been forgotten. This occurs because the dynamics without short-term plasticity transiently moves towards a stored memory before departing away from it. Thus short-term plasticity, operating during the transient, lowers the energy in the vicinity of the stored memory, eventually trapping the combined neuronal and synaptic dynamics at a fixed point close to the stored memory. In this manner, short-term plasticity enables the recall of memories that would otherwise be forgotten, by trapping transients that would otherwise escape. We furthermore find an optimal time constant for short-term synaptic plasticity, matched to the transient dynamics, to empower recall of forgotten memories.

q-bio.NC↗

arXiv2025

The geometry and dynamics of annealed optimization in the coherent Ising machine with hidden and planted solutions

The coherent Ising machine (CIM) is a nonconventional hardware architecture for finding approximate solutions to large-scale combinatorial optimization problems. It operates by annealing a laser gain parameter to adiabatically deform a high-dimensional energy landscape over a set of soft spins, going from a simple convex landscape to the more complex optimization landscape of interest. We address how the evolving energy landscapes guides the optimization dynamics against problems with hidden planted solutions. We study the Sherrington-Kirkpatrick spin-glass with ferromagnetic couplings that favor a hidden configuration by combining the replica method, random matrix theory, the Kac-Rice method and dynamical mean field theory. We characterize energy, number, location, and Hessian eigenspectra of global minima, local minima, and critical points as the landscape evolves. We find that low energy global minima develop soft-modes which the optimization dynamics can exploit to descend the energy landscape. Even when these global minima are aligned to the hidden configuration, there can be exponentially many higher energy local minima that are all unaligned with the hidden solution. Nevertheless, the annealed optimization dynamics can evade this cloud of unaligned high energy local minima and descend near to aligned lower energy global minima. Eventually, as the landscape is further annealed, these global minima become rigid, terminating any further optimization gains from annealing. We further consider a second optimization problem, the Wishart planted ensemble, which contains a hidden planted solution in a landscape with tunable ruggedness. We describe CIM phase transitions between recoverability and non-recoverability of the hidden solution. Overall, we find intriguing relations between high-dimensional geometry and dynamics in analog machines for combinatorial optimization.

cond-mat.dis-nn↗

arXiv2025

CMT-Benchmark: A Benchmark for Condensed Matter Theory Built by Expert Researchers

Large language models (LLMs) have shown remarkable progress in coding and math problem-solving, but evaluation on advanced research-level problems in hard sciences remains scarce. To fill this gap, we present CMT-Benchmark, a dataset of 50 problems covering condensed matter theory (CMT) at the level of an expert researcher. Topics span analytical and computational approaches in quantum many-body, and classical statistical mechanics. The dataset was designed and verified by a panel of expert researchers from around the world. We built the dataset through a collaborative environment that challenges the panel to write and refine problems they would want a research assistant to solve, including Hartree-Fock, exact diagonalization, quantum/variational Monte Carlo, density matrix renormalization group (DMRG), quantum/classical statistical mechanics, and model building. We evaluate LLMs by programmatically checking solutions against expert-supplied ground truth. We developed machine-grading, including symbolic handling of non-commuting operators via normal ordering. They generalize across tasks too. Our evaluations show that frontier models struggle with all of the problems in the dataset, highlighting a gap in the physical reasoning skills of current LLMs. Notably, experts identified strategies for creating increasingly difficult problems by interacting with the LLMs and exploiting common failure modes. The best model, GPT5, solves 30\% of the problems; average across 17 models (GPT, Gemini, Claude, DeepSeek, Llama) is 11.4\pm2.1\%. Moreover, 18 problems are solved by none of the 17 models, and 26 by at most one. These unsolved problems span Quantum Monte Carlo, Variational Monte Carlo, and DMRG. Answers sometimes violate fundamental symmetries or have unphysical scaling dimensions. We believe this benchmark will guide development toward capable AI research assistants and tutors.

cs.LG↗

arXiv2025

Passive objects in a chiral active bath: from micro to macro

We present a detailed derivation of the Langevin dynamics obeyed by a massive rigid body immersed in a chiral active bath. We show how the antisymmetric nature of the noise leads to an unusual relationship between the Langevin equation describing stochastic trajectories and the Fokker-Planck equation governing the time-evolution of the probability density. The chirality of the bath endows the object dynamics with odd diffusivity, odd mobility, and rotational ratchet effects that depend on the object symmetries. For rotationally-symmetric objects, we show that a hidden time-reversal symmetry leads to separate effective equilibrium descriptions for the translational and rotational degrees of freedom. Finally, starting from the bath dynamics, we construct a multipole expansion to quadrupolar order that allows predicting the far-field current and density modulation induced by the object on the bath.

cond-mat.stat-mech↗

arXiv2025

Irreversible swap algorithms for soft sphere glasses

We extend to soft repulsive interaction potentials a recently proposed irreversible swap algorithm originally designed for polydisperse hard spheres. The original algorithm performs rejection-free, irreversible, collective swap moves. We show that event-driven cluster updates of particle diameters can also be performed in continuous potentials by introducing a factorised Metropolis probability. However, the Metropolis factorisation needed to deal with continuous potentials decreases the efficiency of the algorithm and mitigates the benefits of breaking detailed balance. This leads us to propose another irreversible swap algorithm using the standard Metropolis probability that accelerates the relaxation of soft sphere glasses at low temperatures, compared to the original swap algorithm. We apply these efficient swap algorithms to produce very stable inherent structures with vibrational density of states lacking the quasi-localised excitations observed in conventional glasses.

cond-mat.soft↗

arXiv2024

Odd dynamics of passive objects in a chiral active bath

When submerged in a chiral active bath, a passive object becomes a spinning ratchet imbued with odd transport properties. We present the most general Langevin dynamics for a rigid body in a chiral active bath, in the adiabatic limit of large object mass. For rotationally symmetric objects, odd diffusion and odd mobility are connected by an Einstein relation, that we show numerically to break down outside the adiabatic limit. As the object symmetry decreases, its dynamics becomes increasingly irreversible: a massive disk exhibits an effective equilibrium dynamics, while a rod admits distinct translational and rotational temperatures, and a wedge is fully irreversible.Conversely, this departure from equilibrium can be read in universal far-field currents and density modulations of the bath, which we measure numerically and derive analytically.

cond-mat.stat-mech↗

arXiv2024

Clever algorithms for glasses work by time reparametrization

The ultraslow dynamics of glass-formers has been explained by two views considered as mutually exclusive: one invokes locally hindered mobility, the other rests on the complexity of the configuration space. Here we demonstrate that the evolution responds strongly to the details of the dynamics by changing the speed of time-flow: it has time-reparametrization softness. This finding reconciles both views: while local constraints reparametrize the flow of time, the global landscape determines relationships between different correlations at the same times. We show that modern algorithms developed to accelerate the relaxation to equilibrium act by changing the time reparametrization. Their success thus relies on their ability to exploit reparametrization softness. We conjecture that these results extend beyond the realm of glasses to the optimization of more general constraint satisfaction problems and to broader classes of algorithms.

cond-mat.dis-nn↗

arXiv2024

Irreversible Boltzmann samplers in dense liquids: weak-coupling approximation and mode-coupling theory

Exerting a nonequilibrium drive on an otherwise equilibrium Langevin process brings the dynamics out of equilibrium but can also speedup the approach to the Boltzmann steady-state. Transverse forces are a minimal framework to achieve dynamical acceleration of the Boltzmann sampling. We consider a simple liquid in three space dimensions subjected to additional transverse pairwise forces, and quantify the extent to which transverse forces accelerate the dynamics. We first explore the dynamics of a tracer in a weak coupling regime describing high temperatures. The resulting acceleration is correlated with a monotonous increase of the magnitude of odd transport coefficients (mobility and diffusivity) with the amplitude of the transverse drive. We then develop a nonequilibrium version of the mode-coupling theory able to capture the effect of transverse forces, and more generally of forces created by additional degrees of freedom. Based on an analysis of transport coefficients, both odd and longitudinal, both for the collective modes and for a tracer particle, we find a systematic acceleration of the dynamics. Quantitatively, the gain, which is guaranteed throughout the ergodic phase, turns out to be a decreasing function of temperature beyond a temperature crossover, in particular as the glass transition is approached. Our theoretical results are in good agreement with available numerical results.

cond-mat.soft↗

arXiv2024

Transverse forces and glassy liquids in infinite dimensions

We explore the dynamics of a simple liquid whose particles, in addition to standard potential-based interactions, are also subjected to transverse forces preserving the Boltzmann distribution. We derive the effective dynamics of one and two tracer particles in the infinite-dimensional limit. We determine the amount of acceleration of the dynamics caused by the transverse forces, in particular in the vicinity of the glass transition. We analyze the emergence and evolution of odd transport phenomena induced by the transverse forces.

cond-mat.soft↗

arXiv2024

Irreversible Monte Carlo algorithms for hard disk glasses: from event-chain to collective swaps

Equilibrium sampling of the configuration space in disordered systems requires algorithms that bypass the glassy slowing down of the physical dynamics. Irreversible Monte Carlo algorithms breaking detailed balance successfully accelerate sampling in some systems. We first implement an irreversible event-chain Monte Carlo algorithm in a model of polydisperse hard disks. The effect of collective translational moves marginally affects the dynamics and results in a modest speedup that decreases with density. We then propose an irreversible algorithm performing collective particle swaps which outperforms all known Monte Carlo algorithms. We show that these collective swaps can also be used to prepare very dense jammed packings of disks.

cond-mat.dis-nn↗

arXiv2023

Shear-Induced Phase Behavior and Topological Defects in Two-Dimensional Crystals

We investigate through numerical simulations how a two-dimensional crystal yields and flows under an applied shear. We focus over a range that allows us to both address the response in the limit of an infinitesimal shear rate and describe the phase behavior of the system at a finite shear rate. In doing so, we carefully discuss the role of the topological defects and of the finite-size effects. We map out the whole phase diagram of the flowing steady state in the plane formed by temperature and shear rate. Shear-induced melting of the two-dimensional crystal is found to proceed in two steps: first, the solid loses long-range bond-orientational order and flows, even for an infinitesimal shear rate (in the thermodynamic limit). The resulting flowing hexatic phase then melts to a flowing, rather isotropic, liquid at a finite shear rate that depends on temperature. Finally, at a high shear rate, a third regime corresponding to a strongly anisotropic string-like flowing phase appears.

cond-mat.soft↗

arXiv2023

Sampling efficiency of transverse forces in dense liquids

Sampling the Boltzmann distribution using forces that violate detailed balance can be faster than with the equilibrium evolution, but the acceleration depends on the nature of the nonequilibrium drive and the physical situation. Here, we study the efficiency of forces transverse to energy gradients in dense liquids through a combination of techniques: Brownian dynamics simulations, exact infinite-dimensional calculation and a mode-coupling approximation. We find that the sampling speedup varies non-monotonically with temperature, and decreases as the system becomes more glassy. We characterize the interplay between the distance to equilibrium and the efficiency of transverse forces by means of odd transport coefficients.

cond-mat.soft↗

arXiv2022

Accelerating, to some extent, the $p$-spin dynamics

We consider a detailed-balance violating dynamics whose stationary state is a prescribed Boltzmann distribution. Such dynamics have been shown to be faster than any equilibrium counterpart. We quantify the gain in convergence speed for a system whose energy landscape displays one, and then an infinite number of, energy barriers. In the latter case, we work with the mean-field disordered $p$-spin, and show that the convergence to equilibrium or to the nonergodic phase is accelerated, both during the $β$ and $α$-relaxation stages. An interpretation in terms of trajectories in phase space and of an accidental fluctuation-dissipation theorem is provided.

cond-mat.stat-mech↗