arXiv subjects

Yi Sun

Publications and source records attributed to Yi Sun.

At least 55 records · Page 3Linked to original sources

arXiv2025

HiLoMix: Robust High- and Low-Frequency Graph Learning Framework for Mixing Address Association

As mixing services are increasingly being exploited by malicious actors for illicit transactions, mixing address association has emerged as a critical research task. A range of approaches have been explored, with graph-based models standing out for their ability to capture structural patterns in transaction networks. However, these approaches face two main challenges: label noise and label scarcity, leading to suboptimal performance and limited generalization. To address these, we propose HiLoMix, a graph-based learning framework specifically designed for mixing address association. First, we construct the Heterogeneous Attributed Mixing Interaction Graph (HAMIG) to enrich the topological structure. Second, we introduce frequency-aware graph contrastive learning that captures complementary structural signals from high- and low-frequency graph views. Third, we employ weak supervised learning that assigns confidence-based weighting to noisy labels. Then, we jointly train high-pass and low-pass GNNs using both unsupervised contrastive signals and confidence-based supervision to learn robust node representations. Finally, we adopt a stacking framework to fuse predictions from multiple heterogeneous models, further improving generalization and robustness. Experimental results demonstrate that HiLoMix outperforms existing methods in mixing address association.

cs.SI↗

arXiv2025

Dexterous Manipulation Transfer via Progressive Kinematic-Dynamic Alignment

The inherent difficulty and limited scalability of collecting manipulation data using multi-fingered robot hand hardware platforms have resulted in severe data scarcity, impeding research on data-driven dexterous manipulation policy learning. To address this challenge, we present a hand-agnostic manipulation transfer system. It efficiently converts human hand manipulation sequences from demonstration videos into high-quality dexterous manipulation trajectories without requirements of massive training data. To tackle the multi-dimensional disparities between human hands and dexterous hands, as well as the challenges posed by high-degree-of-freedom coordinated control of dexterous hands, we design a progressive transfer framework: first, we establish primary control signals for dexterous hands based on kinematic matching; subsequently, we train residual policies with action space rescaling and thumb-guided initialization to dynamically optimize contact interactions under unified rewards; finally, we compute wrist control trajectories with the objective of preserving operational semantics. Using only human hand manipulation videos, our system automatically configures system parameters for different tasks, balancing kinematic matching and dynamic optimization across dexterous hands, object categories, and tasks. Extensive experimental results demonstrate that our framework can automatically generate smooth and semantically correct dexterous hand manipulation that faithfully reproduces human intentions, achieving high efficiency and strong generalizability with an average transfer success rate of 73%, providing an easily implementable and scalable method for collecting robot dexterous manipulation data.

cs.RO↗

arXiv2025

Bridging the behavior-neural gap: A multimodal AI reveals the brain's geometry of emotion more accurately than human self-reports

The ability to represent emotion plays a significant role in human cognition and social interaction, yet the high-dimensional geometry of this affective space and its neural underpinnings remain debated. A key challenge, the `behavior-neural gap,' is the limited ability of human self-reports to predict brain activity. Here we test the hypothesis that this gap arises from the constraints of traditional rating scales and that large-scale similarity judgments can more faithfully capture the brain's affective geometry. Using AI models as `cognitive agents,' we collected millions of triplet odd-one-out judgments from a multimodal large language model (MLLM) and a language-only model (LLM) in response to 2,180 emotionally evocative videos. We found that the emergent 30-dimensional embeddings from these models are highly interpretable and organize emotion primarily along categorical lines, yet in a blended fashion that incorporates dimensional properties. Most remarkably, the MLLM's representation predicted neural activity in human emotion-processing networks with the highest accuracy, outperforming not only the LLM but also, counterintuitively, representations derived directly from human behavioral ratings. This result supports our primary hypothesis and suggests that sensory grounding--learning from rich visual data--is critical for developing a truly neurally-aligned conceptual framework for emotion. Our findings provide compelling evidence that MLLMs can autonomously develop rich, neurally-aligned affective representations, offering a powerful paradigm to bridge the gap between subjective experience and its neural substrates. Project page: https://reedonepeck.github.io/ai-emotion.github.io/.

cs.HC↗

arXiv2025

High-Precision Temperature Estimation Based on Magnetic Nanoparticles Dominated by Brownian Relaxation under Combined AC and DC Magnetic Fields

Brownian relaxation is one of the primary mechanisms that allows magnetic nanoparticles (MNPs) to convert magnetic energy into thermal energy under an excitation magnetic field. Accurately characterizing the MNPs' magnetization dynamics dominated by Brownian relaxation is crucial for achieving high-precision temperature estimation. However, the lack of a readily applicable analytical expression remains a major obstacle to the advancement of magnetic nanoparticle hyperthermia (MNPH). In this paper, the perturbation method was applied to derive analytical expressions from the Fokker-Planck equation, which characterized MNPs' magnetization behaviors under the AC and DC magnetic fields. Numerical simulations were conducted to validate the accuracy of the analytical expressions and to explore the correlation between temperature and the magnetization response. Then, a temperature analysis model based on magnetization harmonics was constructed. The first and second harmonic ratios and first harmonic phase were used to calculate MNPs' temperature, respectively. The experimental results demonstrate that within 310 K to 320 K, the estimation error of the temperature using the amplitude ratio of the first to second harmonics is below 0.0151 K, while the error using the first harmonic phase is below 0.0218 K. The derived analytical expressions are expected to enhance the accuracy of MNP-based temperature measurements and facilitate their broader applications in MNPH and MNP imaging.

cond-mat.mes-hall↗

arXiv2025

IEFS-GMB: Gradient Memory Bank-Guided Feature Selection Based on Information Entropy for EEG Classification of Neurological Disorders

Deep learning-based EEG classification is crucial for the automated detection of neurological disorders, improving diagnostic accuracy and enabling early intervention. However, the low signal-to-noise ratio of EEG signals limits model performance, making feature selection (FS) vital for optimizing representations learned by neural network encoders. Existing FS methods are seldom designed specifically for EEG diagnosis; many are architecture-dependent and lack interpretability, limiting their applicability. Moreover, most rely on single-iteration data, resulting in limited robustness to variability. To address these issues, we propose IEFS-GMB, an Information Entropy-based Feature Selection method guided by a Gradient Memory Bank. This approach constructs a dynamic memory bank storing historical gradients, computes feature importance via information entropy, and applies entropy-based weighting to select informative EEG features. Experiments on four public neurological disease datasets show that encoders enhanced with IEFS-GMB achieve accuracy improvements of 0.64% to 6.45% over baseline models. The method also outperforms four competing FS techniques and improves model interpretability, supporting its practical use in clinical settings.

cs.LG↗

arXiv2025

BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens

Recent advancements in Large Language Models (LLMs) have leveraged increased test-time computation to enhance reasoning capabilities, a strategy that, while effective, incurs significant latency and resource costs, limiting their applicability in real-world time-constrained or cost-sensitive scenarios. This paper introduces BudgetThinker, a novel framework designed to empower LLMs with budget-aware reasoning, enabling precise control over the length of their thought processes. We propose a methodology that periodically inserts special control tokens during inference to continuously inform the model of its remaining token budget. This approach is coupled with a comprehensive two-stage training pipeline, beginning with Supervised Fine-Tuning (SFT) to familiarize the model with budget constraints, followed by a curriculum-based Reinforcement Learning (RL) phase that utilizes a length-aware reward function to optimize for both accuracy and budget adherence. We demonstrate that BudgetThinker significantly surpasses strong baselines in maintaining performance across a variety of reasoning budgets on challenging mathematical benchmarks. Our method provides a scalable and effective solution for developing efficient and controllable LLM reasoning, making advanced models more practical for deployment in resource-constrained and real-time environments.

cs.LG↗

arXiv2025

LV-CadeNet: A Long-View Feature Convolution-Attention Fusion Encoder-Decoder Network for EEG/MEG Spike Analysis

The analysis of interictal epileptiform discharges (IEDs) in magnetoencephalography (MEG) or electroencephalogram (EEG) recordings represents a critical component in the diagnosis of epilepsy. However, manual analysis of these IEDs, which appear as epileptic spikes, from the large amount of MEG/EEG data is labor intensive and requires high expertise. Although automated methods have been developed to address this challenge, current approaches fail to fully emulate clinical experts' diagnostic intelligence in two key aspects: (1) their analysis on the input signals is limited to short temporal windows matching individual spike durations, missing the extended contextual patterns clinicians use to assess significance; and (2) they fail to adequately capture the dipole patterns with simultaneous positive-negative potential distributions across adjacent sensors that serve as clinicians' key diagnostic criterion for IED identification. To bridge this artificial-human intelligence gap, we propose a novel deep learning framework LV-CadeNet that integrates two key innovations: (1) a Long-View morphological feature representation that mimics expert clinicians' comprehensive assessment of both local spike characteristics and long-view contextual information, and (2) a hierarchical Encoder-Decoder NETwork that employs Convolution-Attention blocks for multi-scale spatiotemporal feature learning with progressive abstraction. Extensive evaluations confirm the superior performance of LV-CadeNet, which outperforms six state-of-the-art methods in EEG spike classification on TUEV, the largest public EEG spike dataset. Additionally, LV-CadeNet attains a significant improvement of 13.58% in balanced accuracy over the leading baseline for MEG spike detection on a clinical MEG dataset from Sanbo Brain Hospital, Capital Medical University.

cs.CV↗

arXiv2025

Human-like object concept representations emerge naturally in multimodal large language models

Understanding how humans conceptualize and categorize natural objects offers critical insights into perception and cognition. With the advent of Large Language Models (LLMs), a key question arises: can these models develop human-like object representations from linguistic and multimodal data? In this study, we combined behavioral and neuroimaging analyses to explore the relationship between object concept representations in LLMs and human cognition. We collected 4.7 million triplet judgments from LLMs and Multimodal LLMs (MLLMs) to derive low-dimensional embeddings that capture the similarity structure of 1,854 natural objects. The resulting 66-dimensional embeddings were stable, predictive, and exhibited semantic clustering similar to human mental representations. Remarkably, the dimensions underlying these embeddings were interpretable, suggesting that LLMs and MLLMs develop human-like conceptual representations of objects. Further analysis showed strong alignment between model embeddings and neural activity patterns in brain regions such as EBA, PPA, RSC, and FFA. This provides compelling evidence that the object representations in LLMs, while not identical to human ones, share fundamental similarities that reflect key aspects of human conceptual knowledge. Our findings advance the understanding of machine intelligence and inform the development of more human-like artificial cognitive systems.

cs.AI↗

arXiv2025

An Empirical Study of LLM Reasoning Ability Under Strict Output Length Constraint

Recent work has demonstrated the remarkable potential of Large Language Models (LLMs) in test-time scaling. By making models think before answering, they are able to achieve much higher accuracy with extra inference computation. However, in many real-world scenarios, models are used under time constraints, where an answer should be given within a certain output length. It is unclear whether and how the reasoning ability of different LLMs remain effective under strict constraints. We take a first look at this problem by conducting an in-depth empirical study. Specifically, we test 30 LLMs on common reasoning datasets under a wide range of output length budgets, and we analyze the correlation between the inference accuracy and various properties including model type, model size, prompt style, etc. We also consider the mappings between token budgets and actual on-device latency budgets. The results have demonstrated several interesting findings regarding the budget-aware LLM reasoning ability that differ from the unconstrained situation, e.g. the optimal choices of either model size or prompt style change under different budgets. These findings offer timely evaluation to this area and practical guidance for users to deploy LLMs under real-world latency constraints.

cs.AI↗

arXiv2025

Learning Task-preferred Inference Routes for Gradient De-conflict in Multi-output DNNs

Multi-output deep neural networks(MONs) contain multiple task branches, and these tasks usually share partial network filters that lead to the entanglement of different task inference routes. Due to the inconsistent optimization objectives, the task gradients used for training MONs will interfere with each other on the shared routes, which will decrease the overall model performance. To address this issue, we propose a novel gradient de-conflict algorithm named DR-MGF(Dynamic Routes and Meta-weighted Gradient Fusion) in this work. Different from existing de-conflict methods, DR-MGF achieves gradient de-conflict in MONs by learning task-preferred inference routes. The proposed method is motivated by our experimental findings: the shared filters are not equally important to different tasks. By designing the learnable task-specific importance variables, DR-MGF evaluates the importance of filters for different tasks. Through making the dominances of tasks over filters be proportional to the task-specific importance of filters, DR-MGF can effectively reduce the inter-task interference. The task-specific importance variables ultimately determine task-preferred inference routes at the end of training iterations. Extensive experimental results on CIFAR, ImageNet, and NYUv2 illustrate that DR-MGF outperforms the existing de-conflict methods both in prediction accuracy and convergence speed of MONs. Furthermore, DR-MGF can be extended to general MONs without modifying the overall network structures.

cs.CV↗

arXiv2025

Full spectral response of grating-induced loss in photonic crystal microrings

Photonic crystal microrings (PhCRs) have emerged as powerful and versatile platforms for integrated nonlinear photonics, offering precise control over frequency and phase matching while maintaining high optical quality factors. Through grating-mediated mode coupling, PhCRs enable advanced dispersion engineering, which is critical for wideband nonlinear processes such as optical parametric oscillation, Kerr frequency comb generation, and dual-pump spontaneous and Bragg scattering four-wave mixing. Beyond dispersion control, PhCRs also facilitate the manipulation of orbital angular momentum (OAM) emission, a key functionality for encoding high-dimensional quantum states in emerging quantum photonic platforms. Despite these advances, the broadband spectral behavior of grating-induced losses in PhCRs remains largely unexplored, with most studies focusing on grating periods near the modal wavelength or its half. Such losses can significantly impact broadband nonlinear processes, where excess loss at unintended wavelengths can degrade device performance. In this work, we experimentally characterize grating-induced losses in PhCRs and reveal their full spectral response as a function of the ratio between modal wavelength and grating period. We identify distinct loss channels arising from either radiation or mode conversion, including a broad excess-loss region attributed to vertical out-coupling into OAM-carrying states. These observations are supported by three-dimensional finite-difference time-domain simulations and further analyzed through OAM radiation angle and phase-mismatch analysis. The resulting broadband loss spectrum highlights critical design trade-offs and provides practical guidelines for optimizing PhCR-based devices for nonlinear photonic applications involving widely separated frequencies.

physics.optics↗

arXiv2025

Parasitic loss in microring-waveguide coupling and its impact on wideband nonlinear photonics

Microring resonators enable the enhancement of nonlinear frequency mixing processes, generating output fields at frequencies that widely differ from the inputs, in some cases by more than an octave. The efficiency of such devices depends on effective in- and out-coupling between access waveguides and the microrings at these widely separated frequencies. One successful approach is to separate the coupling task across multiple waveguides, with a cutoff waveguide (a waveguide that does not support guided modes above a certain wavelength) being judiciously used to prevent unwanted excessive overcoupling at low frequencies. Here, we examine how such a cutoff waveguide can still induce parasitic loss in the coupling region of a microring resonator, thereby impacting nonlinear device performance. We verified this parasitic loss channel through both experiment and simulation, showing that a waveguide optimized for 532 nm (visible) and 780 nm (near-infrared), while nominally cut off at 1550 nm, can still introduce significant parasitic loss at telecom wavelengths. This is studied in the context of visible-telecom optical parametric oscillation, where the excess parasitic loss can be strong enough to prevent threshold from being reached. Our finding elucidates a major challenge for wideband integrated nonlinear photonics processes when efficient coupling of widely-separated frequencies is needed.

physics.optics↗

arXiv2025

Fast reversal of Néel vectors in antiferromagnets via domain-wall motion driven by vertically injected spin currents

In this work, we investigate the steady propagation of 180-degree domain walls (180DWs) of Néel vectors in thin antiferromagnetic strips under perpendicularly injected spin currents with various polarization orientations. Our results show that only spin currents polarized normally to the strip plane guarantee a quick and steady rigid flow of 180DWs, thus realize a fast reversal of Néel vectors in antiferromagnets (AFMs). Different from the common ``current-in-plane'' geometry which is feasible only for metallic AFMs, our ``current-out-of-plane'' layout under investigation can further apply to insulating AFMs (which are more common in real applications) via quantum tunneling effect. Results from this work paves the way for fine control of Néel vectors in (both metallic and insulating) AFM strips and further development of novel magnetic nanodevices based on them.

cond-mat.mes-hall↗

arXiv2025

Tightly-confined and long Z-cut lithium niobate waveguide with ultralow-loss

Lithium niobate (LN) is a promising material for future complex photonic-electronic circuits, with wide applications in fields like data communications, sensing, optical computation, and quantum optics. There was a great step toward LN photonic integrated circuits (PICs) with the development of dry etching for low-loss LN on insulator (LNOI) waveguides. However, the versatility of the LN waveguide platform for applications like $χ^3$ nonlinear devices and passive phase sensitive components, has not been fully utilized. The main challenges are the difficulty of making highly confined ultralow-loss waveguides and overcoming the strong material birefringence. Here, we developed a fabrication technology for an ultralow-loss, tightly-confined, dispersion-engineered LN waveguide. We demonstrated an ultra-low propagation loss of 5.8 dB/m in a decimeter-long LN spiral waveguide. We focused on Z-cut LN waveguides with TE mode to avoid the material birefringence. Aiming for $χ^3$ nonlinear applications, we demonstrated the first all normal-dispersion (ANDi) based coherent octave-spanning supercontinuum frequency comb in integrated LN waveguide. Our ultralow-loss Z-cut LN long waveguide might be useful in on-chip narrow linewidth lasers, optical delay lines, and parametric amplifiers.

physics.optics↗

arXiv2025

Automated Detection of Epileptic Spikes and Seizures Incorporating a Novel Spatial Clustering Prior

A Magnetoencephalography (MEG) time-series recording consists of multi-channel signals collected by superconducting sensors, with each signal's intensity reflecting magnetic field changes over time at the sensor location. Automating epileptic MEG spike detection significantly reduces manual assessment time and effort, yielding substantial clinical benefits. Existing research addresses MEG spike detection by encoding neural network inputs with signals from all channel within a time segment, followed by classification. However, these methods overlook simultaneous spiking occurred from nearby sensors. We introduce a simple yet effective paradigm that first clusters MEG channels based on their sensor's spatial position. Next, a novel convolutional input module is designed to integrate the spatial clustering and temporal changes of the signals. This module is fed into a custom MEEG-ResNet3D developed by the authors, which learns to extract relevant features and classify the input as a spike clip or not. Our method achieves an F1 score of 94.73% on a large real-world MEG dataset Sanbo-CMR collected from two centers, outperforming state-of-the-art approaches by 1.85%. Moreover, it demonstrates efficacy and stability in the Electroencephalographic (EEG) seizure detection task, yielding an improved weighted F1 score of 1.4% compared to current state-of-the-art techniques evaluated on TUSZ, whch is the largest EEG seizure dataset.

eess.SP↗

arXiv2024

The Codecfake Dataset and Countermeasures for the Universally Detection of Deepfake Audio

With the proliferation of Audio Language Model (ALM) based deepfake audio, there is an urgent need for generalized detection methods. ALM-based deepfake audio currently exhibits widespread, high deception, and type versatility, posing a significant challenge to current audio deepfake detection (ADD) models trained solely on vocoded data. To effectively detect ALM-based deepfake audio, we focus on the mechanism of the ALM-based audio generation method, the conversion from neural codec to waveform. We initially constructed the Codecfake dataset, an open-source, large-scale collection comprising over 1 million audio samples in both English and Chinese, focus on ALM-based audio detection. As countermeasure, to achieve universal detection of deepfake audio and tackle domain ascent bias issue of original sharpness aware minimization (SAM), we propose the CSAM strategy to learn a domain balanced and generalized minima. In our experiments, we first demonstrate that ADD model training with the Codecfake dataset can effectively detects ALM-based audio. Furthermore, our proposed generalization countermeasure yields the lowest average equal error rate (EER) of 0.616% across all test conditions compared to baseline models. The dataset and associated code are available online.

cs.SD↗

arXiv2024

ChainStream: An LLM-based Framework for Unified Synthetic Sensing

Many applications demand context sensing to offer personalized and timely services. Yet, developing sensing programs can be challenging for developers and using them is privacy-concerning for end-users. In this paper, we propose to use natural language as the unified interface to process personal data and sense user context, which can effectively ease app development and make the data pipeline more transparent. Our work is inspired by large language models (LLMs) and other generative models, while directly applying them does not solve the problem - letting the model directly process the data cannot handle complex sensing requests and letting the model write the data processing program suffers error-prone code generation. We address the problem with 1) a unified data processing framework that makes context-sensing programs simpler and 2) a feedback-guided query optimizer that makes data query more informative. To evaluate the performance of natural language-based context sensing, we create a benchmark that contains 133 context sensing tasks. Extensive evaluation has shown that our approach is able to automatically solve the context-sensing tasks efficiently and precisely. The code is opensourced at https://github.com/MobileLLM/ChainStream.

cs.CL↗

arXiv2024

Silicon photonic MEMS switches based on split waveguide crossings

The continuous push for high-performance photonic switches is one of the most crucial premises for the sustainable scaling of programmable and reconfigurable photonic circuits for a wide spectrum of applications. Conventional optical switches rely on the perturbative mechanisms of mode coupling or mode interference, resulting in inherent bottlenecks in their switching performance concerning size, power consumption and bandwidth. Here we propose and realize a silicon photonic 2x2 elementary switch based on a split waveguide crossing (SWX) consisting of two halves. The propagation direction of the incident light is manipulated to implement the OFF/ON states by splitting/combining the two halves of the SWX, showing excellent performance with low excess loss and low crosstalk over an ultrawide bandwidth. Both elementary switch and a 64x64 switch array based on Benes topology are fabricated and characterized, demonstrating great potential for practical scenarios such as photonic interconnect/routing, Lidar and spectroscopy, photonic computing, as well as microwave photonics.

physics.optics↗