arXiv subjects

Fei Song

Publications and source records attributed to Fei Song.

At least 19 recordsLinked to original sources

arXiv2026

Strain-Tuned Nodal Superconductivity in the Charge-Ordered Kagome Metal CsV$_3$Sb$_5$

The nature of the superconducting pairing symmetry in the kagome metal CsV$_3$Sb$_5$ and its relationship with the charge density wave (CDW) order are central unresolved issues. Here, we investigate the evolution of superconductivity in CsV$_3$Sb$_5$ under in-situ uniaxial pressure using $^{121}$Sb nuclear quadrupole resonance (NQR). We find that tensile strain significantly enhances the superconducting transition temperature, $T_{\rm c}$, while the CDW remains unchanged, demonstrating that superconductivity can be tuned independently of the bulk charge order. At a tensile strain of $\varepsilon$ = +0.90%, the nuclear spin-lattice relaxation rate reveals a remarkable double transition: an upper transition at $T_{\rm c1}$ = 3.6 K to a nodal gap state, and a lower one at $T_{\rm c2}$ = 3.0 K characterized by a nodeless gap. These results evidence degenerate superconducting states with different gap symmetry in the kagome metal at ambient pressure which split under strain. Our work demonstrates a high tunability of superconductivity by uniaxial pressure.

cond-mat.supr-con↗

arXiv2026

Orchestrating Data Collection and Computation in Green IoT Networks

Future Internet of things (IoT) networks will host applications that involve data collection and computation tasks on one or more servers. To this end, this paper proposes the first mixed integer linear program (MILP) to schedule and embed applications on energy harvesting nodes, where it optimizes (i) the sampling time of devices, (ii) whether to run an application, and (iii) the energy usage of devices, gateways and servers. To ensure applications are run often, we adopt the maximum age of service (AoS) metric, and set the MILP's objective to minimize the maximum AoS or min-max AoS of applications. This paper also proposes two novel solutions: (i) a receding horizon control (RHC) based method, and (ii) a solution that greedily embeds applications according to their AoS. The results show that the min-max AoS of RHC and greedy approach is respectively 1.07x and 1.13x higher than MILP.

cs.NI↗

arXiv2026

Boundary-sensitive non-Hermiticity of Floquet Hamiltonian: spectral transition and scale-free localization

We report a novel mechanism of boundary-sensitive PT symmetry breaking in one-dimensional Floquet systems. By designing a time-periodic driving protocol, we realize a Floquet Hamiltonian that is Hermitian under periodic boundary conditions yet acquires non-Hermitian boundary terms under open boundary conditions due to the non-commutativity of driving Hamiltonians. We establish that a PT symmetry breaking transition occurs when the quasienergy bandwidth expands to cover the entire frequency Brillouin zone. This condition highlights a crucial difference from static non-Hermitian systems, where such transitions typically require band touching. Furthermore, we demonstrate that in the PT-broken phase, the eigenstates exhibit scale-free localization, a phenomenon arising from the specific system-size scaling of non-Hermitian terms. Finally, we provide a general framework for constructing multi-band models that exhibit this boundary-induced phase transition.

quant-ph↗

arXiv2026

Stable boundary modes for fragile topology from spontaneous PT-symmetry breaking

Two-dimensional topological insulators protected by nonlocal symmetries or with fragile topology usually do not admit robust in-gap edge modes due to the incompatibility between the symmetry and the boundary. Here, we show that in a parity-time (PT) symmetric system robust in-gap topological edge modes can be stably induced by non-Hermitian couplings that spontaneously break the PT symmetry of the eigenstates. The topological edge modes traverse the imaginary spectral gap between a pair of fragile topological bands, which is opened by the presence of the non-Hermitian perturbation. We demonstrate that the net number of resulting in-gap modes is protected by an operator version of anomaly cancellation that extends beyond the Hermitian limit. The results imply that loss and gain can in principle drive fragile topological phenomena to stable topological phenomena.

cond-mat.mes-hall↗

arXiv2025

AmPLe: Supporting Vision-Language Models via Adaptive-Debiased Ensemble Multi-Prompt Learning

Multi-prompt learning methods have emerged as an effective approach for facilitating the rapid adaptation of vision-language models to downstream tasks with limited resources. Existing multi-prompt learning methods primarily focus on utilizing various meticulously designed prompts within a single foundation vision-language model to achieve superior performance. However, the overlooked model-prompt matching bias hinders the development of multi-prompt learning, i.e., the same prompt can convey different semantics across distinct vision-language models, such as CLIP-ViT-B/16 and CLIP-ViT-B/32, resulting in inconsistent predictions of identical prompt. To mitigate the impact of this bias on downstream tasks, we explore an ensemble learning approach to sufficiently aggregate the benefits of diverse predictions. Additionally, we further disclose the presence of sample-prompt matching bias, which originates from the prompt-irrelevant semantics encapsulated in the input samples. Thus, directly utilizing all information from the input samples for generating weights of ensemble learning can lead to suboptimal performance. In response, we extract prompt-relevant semantics from input samples by leveraging the guidance of the information theory-based analysis, adaptively calculating debiased ensemble weights. Overall, we propose Adaptive-Debiased Ensemble MultiPrompt Learning, abbreviated as AmPLe, to mitigate the two types of bias simultaneously. Extensive experiments on three representative tasks, i.e., generalization to novel classes, new target datasets, and unseen domain shifts, show that AmPLe can widely outperform existing methods. Theoretical validation from a causal perspective further supports the effectiveness of AmPLe.

cs.CV↗

arXiv2025

Doubly Debiased Test-Time Prompt Tuning for Vision-Language Models

Test-time prompt tuning for vision-language models has demonstrated impressive generalization capabilities under zero-shot settings. However, tuning the learnable prompts solely based on unlabeled test data may induce prompt optimization bias, ultimately leading to suboptimal performance on downstream tasks. In this work, we analyze the underlying causes of prompt optimization bias from both the model and data perspectives. In terms of the model, the entropy minimization objective typically focuses on reducing the entropy of model predictions while overlooking their correctness. This can result in overconfident yet incorrect outputs, thereby compromising the quality of prompt optimization. On the data side, prompts affected by optimization bias can introduce misalignment between visual and textual modalities, which further aggravates the prompt optimization bias. To this end, we propose a Doubly Debiased Test-Time Prompt Tuning method. Specifically, we first introduce a dynamic retrieval-augmented modulation module that retrieves high-confidence knowledge from a dynamic knowledge base using the test image feature as a query, and uses the retrieved knowledge to modulate the predictions. Guided by the refined predictions, we further develop a reliability-aware prompt optimization module that incorporates a confidence-based weighted ensemble and cross-modal consistency distillation to impose regularization constraints during prompt tuning. Extensive experiments across 15 benchmark datasets involving both natural distribution shifts and cross-datasets generalization demonstrate that our method outperforms baselines, validating its effectiveness in mitigating prompt optimization bias.

cs.LG↗

arXiv2025

The Principle of Isomorphism: A Theory of Population Activity in Grid Cells and Beyond

Neural population activity organizes into low-dimensional manifolds embedded within high-dimensional state spaces, yet the principles governing the topology and geometry of these manifolds remain elusive. Here, we propose the Principle of Isomorphism (PIso), which posits that the topology of a neural manifold is constrained by the mathematical structure of the computational task it supports. We apply this framework to the mammalian grid cell system through two distinct theoretical lenses: an intrinsic neural metric, which requires a locally flat Riemannian structure, and path integration, which requires a compact connected Abelian Lie group structure. We show that these two routes are both sufficient conditions that converge on the same toroidal latent topology, and that they naturally unify within Euclidean space. Using a minimal feedforward network that constrains population activity to a torus with tunable geometry, we find that hexagonal grid fields emerge only in an intermediate geometric regime, becoming diffuse or square-like otherwise. Our work clarifies the separation between three notions: latent topology, extrinsic embedding geometry, and decoded physical geometry, and identifies the topology of the population code as the more invariant consequence of the task structure, while leaving the precise mechanism that selects hexagonal single-cell firing patterns as an open problem.

q-bio.NC↗

arXiv2025

Non-Bloch edge dynamics of non-Hermitian lattices

The non-Hermitian skin effect, i.e., the localization of nominally bulk modes, not only drastically reshapes the spectral properties of non-Hermitian systems, but also dramatically modifies the real-time dynamics therein. Here we investigate the time evolution of waves (or quantum-mechanical particles) initialized around the edge of non-Hermitian lattices. The non-Hermitian skin effect tends to localize the wave to the edge, meaning that the real-time dynamics differs from the Bloch-theory picture. We focus on the long-time decay or growth rate of wave function, which is quantified by the Lyapunov exponents. These exponents can be obtained from the saddle points in the complex momentum space. We propose an efficient yet unambiguous criterion for identifying the dominant saddle point that determines the Lyapunov exponents. Our criterion can be precisely formulated in terms of a mathematical concept known as the Lefschetz thimble. Counterintuitively, the seemingly natural criterion based on the imaginary part of the energy fails. Our work provides a coherent theory for characterizing the real-time edge dynamics of non-Hermitian lattices. Our predictions are testable in various non-Hermitian physical platforms.

quant-ph↗

arXiv2025

Measurement and Modeling on Terahertz Channel Propagation Through Vegetation

The terahertz band offers promising opportunities for high-capacity wireless communications but faces significant challenges from vegetation-induced channel impairments. This article presents a comprehensive investigation of THz channel propagation through vegetation, introducing a hybrid modeling approach that combines deterministic vegetation dependent exponential decay modeling with statistical characterization of temporal variations. Through extensive laboratory measurements using Epipremnum aureum, we find that vegetation introduces angular-dependent power losses, with channel statistics following heavy tailed Stable distributions rather than conventional Rician or Weibull models. Our outdoor measurements with dense and sparse lilac scenarios reveal pronounced seasonal variations in attenuation and height-dependent effects, while validating the VED model's ability to maintain excellent agreement with measured data and parameter stability across different heights. Critical bit error rate analysis uncovers distinct SNR thresholds beyond which performance exhibits oscillatory behavior due to heavy-tailed fading, with significant implications for modulation scheme selection and power control strategies in practical THz communication systems.

physics.app-ph↗

arXiv2024

Terahertz channel power and BER performance in rain

Terahertz (THz) communications have emerged as a promising technology for 6G networks due to their potential for achieving terabit-per-second data rates. However, the impact of rainfall on THz channel characteristics remains incompletely understood, particularly regarding power attenuation mechanisms and bit error rate (BER) performance. This article presents a systematic measurement-based and theoretical investigation of line-of-sight (LoS) THz channel behavior under rainfall conditions, methodically examining both power attenuation mechanisms and bit error rate (BER) performance. Our experimental campaign, conducted at frequencies of 220-230 GHz over a 54-meter outdoor channel, is complemented by analytical frameworks incorporating ITU-R and Mie scattering models. The study reveals that while rain induces significant power attenuation, multipath scattering effects remain minimal, with Rician K-factors maintaining high values. Notably, we observe substantial variations in power loss under constant rain rates, attributed to dynamic changes in raindrop size distribution. Comparative analysis demonstrates superior BER performance of Quadrature Amplitude Modulation (QAM) in rainfall conditions, while revealing increased environmental sensitivity at higher frequencies. These findings underscore the necessity for adaptive modulation schemes and strategic frequency planning in future THz communication systems.

physics.app-ph↗

arXiv2024

Dynamic Trajectory and Power Control in Ultra-Dense UAV Networks: A Mean-Field Reinforcement Learning Approach

In ultra-dense unmanned aerial vehicle (UAV) networks, it is challenging to coordinate the resource allocation and interference management among large-scale UAVs, for providing flexible and efficient service coverage to the ground users (GUs). In this paper, we propose a learning-based resource allocation scheme in an ultra-dense UAV communication network, where the GUs' service demands are time-varying with unknown distributions. We formulate the non-cooperative game among multiple co-channel UAVs as a stochastic game, where each UAV jointly optimizes its trajectory, user association, and downlink power control to maximize the expectation of its locally cumulative energy efficiency under the interference and energy constraints. To cope with the scalability issue in a large-scale network, we further formulate the problem as a mean-field game (MFG), which simplifies the interactions among the UAVs into a two-player game between a representative UAV and a mean-field. We prove the existence and uniqueness of the equilibrium for the MFG, and propose a model-free mean-field reinforcement learning algorithm named maximum entropy mean-field deep Q network (ME-MFDQN) to solve the mean-field equilibrium in both fully and partially observable scenarios. The simulation results reveal that the proposed algorithm improves the energy efficiency compared with the benchmark algorithms. Moreover, the performance can be further enhanced if the GUs' service demands exhibit higher temporal correlation or if the UAVs have wider observation capabilities over their nearby GUs.

eess.SY↗

arXiv2024

Fragile non-Bloch spectrum and unconventional Green's function

In non-Hermitian systems, it is a counterintuitive feature of the non-Hermitian skin effect (NHSE) that the energy spectrum and eigenstates can be totally different under open or periodic boundary conditions, suggesting that non-Hermitian spectra can be extremely sensitive to non-local perturbations. Here, we show that a wide range of non-Hermitian models with NHSE can even be highly sensitive to local perturbation under open boundary conditions. The spectrum of these models is so fragile that it can be significantly modified by adding only exponentially small perturbations on boundaries. Intriguingly, we show that such fragile spectra are quantified by the Green's function exhibiting unconventional V-shape asymptotic behaviors. Accordingly, bi-directional exponential amplification can be observed. As an interesting consequence, we find a real-to-complex transition of the bulk spectrum induced by exponentially small boundary perturbations. Finally, we reveal a hierarchy of the asymptotic behaviors of non-Hermitian Green's functions, which restricts the frequency range for the presence of unconventional Green's functions.

quant-ph↗

arXiv2024

Many-body Liouvillian dynamics with a non-Hermitian tensor-network kernel polynomial algorithm

Understanding the dynamics of open quantum many-body systems is a major problem in quantum matter. Specifically, efficiently solving the spectrum of the Liouvillian superoperator governing such dynamics remains a critical open challenge. Here, we put forward a method for solving the many-body Liouvillian spectrum and dynamics based on the non-Hermitian kernel polynomial method and tensor-network techniques. We demonstrate the faithfulness of our method by computing the dynamics of the dephasing quantum compass model with a gradient magnetic field and comparing it with exact results. In particular, we show that our method allows us to characterize the quantum Zeno crossover and the reduction of relaxation rate due to Stark localization in this model. We further demonstrate the ability of our method to go beyond exact results by exploring nearest-neighbor interaction effects on the Liouvillian dynamics, elucidating the interplay between Stark localization and many-body interactions. Our method provides an efficient solution to many-body Liouvillian spectrum and dynamics, establishing a methodology to explore large open quantum many-body systems.

quant-ph↗

arXiv2024

Interventional Imbalanced Multi-Modal Representation Learning via $\beta$-Generalization Front-Door Criterion

Multi-modal methods establish comprehensive superiority over uni-modal methods. However, the imbalanced contributions of different modalities to task-dependent predictions constantly degrade the discriminative performance of canonical multi-modal methods. Based on the contribution to task-dependent predictions, modalities can be identified as predominant and auxiliary modalities. Benchmark methods raise a tractable solution: augmenting the auxiliary modality with a minor contribution during training. However, our empirical explorations challenge the fundamental idea behind such behavior, and we further conclude that benchmark approaches suffer from certain defects: insufficient theoretical interpretability and limited exploration capability of discriminative knowledge. To this end, we revisit multi-modal representation learning from a causal perspective and build the Structural Causal Model. Following the empirical explorations, we determine to capture the true causality between the discriminative knowledge of predominant modality and predictive label while considering the auxiliary modality. Thus, we introduce the $\beta$-generalization front-door criterion. Furthermore, we propose a novel network for sufficiently exploring multi-modal discriminative knowledge. Rigorous theoretical analyses and various empirical evaluations are provided to support the effectiveness of the innate mechanism behind our proposed method.

cs.LG↗

arXiv2024

Terahertz channel modeling based on surface sensing characteristics

The dielectric properties of environmental surfaces, including walls, floors and the ground, etc., play a crucial role in shaping the accuracy of terahertz (THz) channel modeling, thereby directly impacting the effectiveness of communication systems. Traditionally, acquiring these properties has relied on methods such as terahertz time-domain spectroscopy (THz-TDS) or vector network analyzers (VNA), demanding rigorous sample preparation and entailing a significant expenditure of time. However, such measurements are not always feasible, particularly in novel and uncharacterized scenarios. In this work, we propose a new approach for channel modeling that leverages the inherent sensing capabilities of THz channels. By comparing the results obtained through channel sensing with that derived from THz-TDS measurements, we demonstrate the method's ability to yield dependable surface property information. The application of this approach in both a miniaturized cityscape scenario and an indoor environment has shown consistency with experimental measurements, thereby verifying its effectiveness in real-world settings.

physics.app-ph↗

arXiv2024

Phase Transition of single-layer vanadium diselenide on Au(111) with distinguished electronic structures

Herein, we report the reversible structural transition of single-layer VSe2 grown on Au(111) through alternating thermal annealing and Se replenishment. Using scanning tunneling microscopy (STM) and angle-resolved photoemission spectroscopy (ARPES), we demonstrate the epitaxial growth of high-quality VSe2 on Au(111) with the octahedral (1T) structure and the Se-vacancy-induced transformation of VSe2 from the metallic moir\'e (1T) phase to the semiconducting (2H) phase. With convincing agreement between the experimental results and DFT calculations, the nanostructure near the grain boundary in the defective intermediate phase is confirmed, as well as the reaction pathway with Se gradually depleting at elevated temperatures. Importantly, it is revealed that the density of the linear Se defects plays a crucial role in the formation of the 2H domain phase due to the increment of the in-plane lattice parameter after Se desorption and the better thermal stability of the 2H phase compared to the 1T phase. The proper control of the density of Se atoms in the topmost Se layer of VSe2 could feasibly manipulate the ratio between the 1T phase and the 2H phase in the steak-shaped domain, which is regarded as a good platform for 2D homojunctions in nanoelectronics.

cond-mat.mtrl-sci↗

arXiv2024

BayesPrompt: Prompting Large-Scale Pre-Trained Language Models on Few-shot Inference via Debiased Domain Abstraction

As a novel and effective fine-tuning paradigm based on large-scale pre-trained language models (PLMs), prompt-tuning aims to reduce the gap between downstream tasks and pre-training objectives. While prompt-tuning has yielded continuous advancements in various tasks, such an approach still remains a persistent defect: prompt-tuning methods fail to generalize to specific few-shot patterns. From the perspective of distribution analyses, we disclose that the intrinsic issues behind the phenomenon are the over-multitudinous conceptual knowledge contained in PLMs and the abridged knowledge for target downstream domains, which jointly result in that PLMs mis-locate the knowledge distributions corresponding to the target domains in the universal knowledge embedding space. To this end, we intuitively explore to approximate the unabridged target domains of downstream tasks in a debiased manner, and then abstract such domains to generate discriminative prompts, thereby providing the de-ambiguous guidance for PLMs. Guided by such an intuition, we propose a simple yet effective approach, namely BayesPrompt, to learn prompts that contain the domain discriminative information against the interference from domain-irrelevant knowledge. BayesPrompt primitively leverages known distributions to approximate the debiased factual distributions of target domains and further uniformly samples certain representative features from the approximated distributions to generate the ultimate prompts for PLMs. We provide theoretical insights with the connection to domain adaptation. Empirically, our method achieves state-of-the-art performance on benchmarks.

cs.CL↗

arXiv2023

High-speed surface-property recognition by 140-GHz frequency

In the field of integrated sensing and communication, there's a growing need for advanced environmental perception. The terahertz (THz) frequency band, significant for ultra-high-speed data connections, shows promise in environmental sensing, particularly in detecting surface textures crucial for autonomous system's decision-making. However, traditional numerical methods for parameter estimation in these environments struggle with accuracy, speed, and stability, especially in high-speed scenarios like vehicle-to-everything communications. This study introduces a deep learning approach for identifying surface roughness using a 140-GHz setup tailored for high-speed conditions. A high-speed data acquisition system was developed to mimic real-world scenarios, and a diverse set of rough surface samples was collected for realistic high-speed datasets to train the models. The model was trained and validated in three challenging scenarios: random occlusions, sparse data, and narrow-angle observations. The results demonstrate the method's effectiveness in high-speed conditions, suggesting terahertz frequencies' potential in future sensing and communication applications.

physics.app-ph↗