Searcharxiv⌕ Search

arXiv subjects

Xin Xue

Publications and source records attributed to Xin Xue.

12 recordsLinked to original sources
arXiv2026

NEST: Tackling Dataset-Level Distribution Shifts via Regime-Oriented Mixture-of-Experts

Accurate long-term forecasting in complex systems is frequently compromised by dataset-level distribution shifts, where diverse underlying behavioral modes and evolving system states drive the dynamic multivariate time-series. While existing methods predominantly focus on local temporal shifts, they fail to explicitly model the global structural challenge where datasets are composites of distinct operational regimes. In this paper, we propose NEST, a specialized framework designed to model and recompose these evolving structures through a two-phase dense MoE architecture. NEST first facilitates structural specialization by partitioning the dataset into distinct operational regimes through unsupervised clustering in a principled moment-entropy space. We introduce a regime-oriented router mechanism that generates initial expert weights based on temporal content, subsequently refined through geometric modulation to regime centroids. Crucially, rather than acting as monolithic predictors, individual experts function as specialized kernels that capture regime-specific dynamics by evolving unique variate-attention patterns. Extensive evaluations on diverse benchmarks, including heterogeneous network traffic and physical phenomena, demonstrate that NEST consistently achieves state-of-the-art performance. Our code and datasets are available at https://github.com/Aaralshin/NEST

cs.LG↗
arXiv2025

Instructing the Architecture Search for Spatial-temporal Sequence Forecasting with LLM

Spatial-temporal sequence forecasting (STSF) is a long-standing research problem with widespread real-world applications. Neural architecture search (NAS), which automates the neural network design, has been shown effective in tackling the STSF problem. However, the existing NAS methods for STSF focus on generating architectures in a time-consuming data-driven fashion, which heavily limits their ability to use background knowledge and explore the complicated search trajectory. Large language models (LLMs) have shown remarkable ability in decision-making with comprehensive internal world knowledge, but how it could benefit NAS for STSF remains unexplored. In this paper, we propose a novel NAS method for STSF based on LLM. Instead of directly generate architectures with LLM, We inspire the LLM's capability with a multi-level enhancement mechanism. Specifically, on the step-level, we decompose the generation task into decision steps with powerful prompt engineering and inspire LLM to serve as instructor for architecture search based on its internal knowledge. On the instance-level, we utilize a one-step tuning framework to quickly evaluate the architecture instance and a memory bank to cumulate knowledge to improve LLM's search ability. On the task-level, we propose a two-stage architecture search, balancing the exploration stage and optimization stage, to reduce the possibility of being trapped in local optima. Extensive experimental results demonstrate that our method can achieve competitive effectiveness with superior efficiency against existing NAS methods for STSF.

cs.CL↗
arXiv2025

Physics-inspired Machine Learning for Quantum Error Mitigation

Noise is a major obstacle in current quantum computing, and Machine Learning for Quantum Error Mitigation (ML-QEM) promises to address this challenge, enhancing computational accuracy while reducing the sampling overheads of standard QEM methods. Yet, existing models lack physical interpretability and rely heavily on extensive datasets, hindering their scalability in large-scale quantum circuits. To tackle these issues, we introduce the Neural Noise Accumulation Surrogate (NNAS), a physics-inspired neural network for ML-QEM that incorporates the structural characteristics of quantum noise accumulation within multi-layer circuits, endowing the model with physical interpretability. Experimental results demonstrate that NNAS outperforms current methods across a spectrum of metrics, including error mitigation capability, quantum resource consumption, and training dataset size. Notably, for deeper circuits where QEM methods typically struggle, NNAS achieves a remarkable reduction of over half in errors. NNAS also demands substantially fewer training data, reducing dataset reliance by at least an order of magnitude, due to its ability to rapidly capture noise accumulation patterns across circuit layers. This work pioneers the integration of quantum process-derived structural characteristics into neural network architectures, broadly enhancing QEM's performance and applicability, and establishes an integrative paradigm that extends to various quantum-inspired neural network architectures.

quant-ph↗
arXiv2013

Trapping Quantum Coherence in Local Energy Minima

Clusters of solid-state quantum devices have long-living metastable states of local energy minima which may be used to store quantum information. The low to vanishing rate of dissipation fulfils the prerequisite to maintain quantum coherence. Then physical symmetrization of the devices could minimize the couplings of the clusters to environmental degrees of freedom so to reduce the rate of decoherence. Combined with various other error correction mechanisms and methods, such designs and optimizations could render solid-state devices useful for quantum information processing, which have the advantages of flexibility in state manipulation and system scaling.

quant-ph↗
arXiv1998

Single Molecule Magnetic Resonance and Quantum Computation

It is proposed that nuclear (or electron) spins in a trapped molecule would be well isolated from the environment and the state of each spin can be measured by means of mechanical detection of magnetic resonance. Therefore molecular traps make an entirely new approach possible for spin-resonance quantum computation which can be conveniently scaled up. In the context of magnetic resonance spectroscopy, a molecular trap promises the ultimate sensitivity for single spin detection and an unprecedented spectral resolution as well.

quant-ph↗
arXiv1998

NMR Quantum Automata in Doped Crystals

In a lattice ${\cal L}$ of nuclear spins with ABCABCABC... type periodic structure embedded in a single-crystal solid, each ABC-unit can be used to store quantum information and the information can be moved around via some cellular shifting mechanism. Impurity doping marks a special site D$\not\in{\cal L}$ which together with the local spin lattice constitute a quantum automaton where the D site serves as the input/output port and universal quantum logic is done through two-body interactions between two spins at D and a nearby site. The novel NMR quantum computer can be easily scaled up and may work at low temperature to overcome the problem of exponential decay in signal-to-noise ratio in room temperature NMR.

quant-ph↗
arXiv1998

Quantum Isomorphic Simulation

A new type of quantum simulator is proposed which can simulate any quantum many-body system in an isomorphic manner. It can actually synthesize a duplicate of the system to be simulated. The isomorphic simulation has the great advantage that the inevitable coupling of the simulator to the environment can be fully exploited in simulating thermodynamic processes.

quant-ph↗
arXiv1998

Static Quantum Computation

Tailoring many-body interactions among a proper quantum system endows it with computing ability by means of static quantum computation in the sense that some of the physical degrees of freedom can be used to store binary information and the corresponding binary variables satisfy some given logic relations if and only if the system is in the ground state. Two theorems are proved showing that the universal static quantum computer can encode the solutions for any P and NP problem into its ground state using only polynomial number (in the problem input size) of logic gates. The second step is to read out the solutions by relaxing the system. The time complexity is relevant when one tries to read out the solution by relaxing the system, therefore our model of static quantum computation provides a new connection between the computational complexity and the dynamics of a complex system.

quant-ph↗
arXiv1997

Tailoring Many-Body Interactions to Solve Hard Combinatorial Problems

A quantum machine consisting of interacting linear clusters of atoms is proposed for the 3SAT problem. Each cluster with two relevant states of collective motion can be used to register a Boolean variable. Given any 3SAT Boolean formula the interactions among the clusters can be so tailored that the ground state(s) (possibly degenerate) of the whole system encodes the satisfying truth assignment(s) for it. This relates the 3SAT problem to the dynamics of the properly designed glass system.

quant-ph↗
arXiv1997

Superconductive Static Quantum Logic

Superconducting rings with exactly $Φ_0/2$ magnetic flux threading are analogous of Ising spins having two degenerate states which can be used to store binary information. When brought close these rings interact by means of magnetic coupling. If the interactions are properly tailored, a system of such superconducting rings can accomplish static quantum logic in the sense that the states of the rings interpreted as Boolean variables satisfy the desired logic relations when and only when the whole system is in the ground state. Such static logic is essential to carry out the static quantum computation [1,2].

quant-ph↗