arXiv subjects

Siyuan Wang

Publications and source records attributed to Siyuan Wang.

At least 73 records · Page 4Linked to original sources

arXiv2025

Response Analysis of Four-Level Heterodyne Rydberg Atom Receiver

The four-level heterodyne Rydberg atom receiver has garnered significant attention in microwave detection and communication due to its high sensitivity and phase measurement capabilities. Existing theoretical studies, primarily based on static solutions, are limited in characterizing the system's frequency response. To address this, this paper comprehensively investigates the dynamic solutions of the density matrix elements for the four-level heterodyne structure, establishing a quantitative relationship between system response, signal frequency, and system parameters. This enables theoretical bandwidth calculations and performance analysis. This paper also constructs a noise model for the density matrix elements, revealing the relationship between the ultimate sensitivity of the Rydberg atom receiver and the noise in the density matrix elements. Both theoretical simulation and experimental results demonstrate that the bandwidth of the four-level heterodyne receiver can exceed 10 MHz. This study provides critical theoretical support for the engineering applications and performance optimization of heterodyne Rydberg atom receivers.

physics.atom-ph↗

arXiv2025

Prompt-guided Disentangled Representation for Action Recognition

Action recognition is a fundamental task in video understanding. Existing methods typically extract unified features to process all actions in one video, which makes it challenging to model the interactions between different objects in multi-action scenarios. To alleviate this issue, we explore disentangling any specified actions from complex scenes as an effective solution. In this paper, we propose Prompt-guided Disentangled Representation for Action Recognition (ProDA), a novel framework that disentangles any specified actions from a multi-action scene. ProDA leverages Spatio-temporal Scene Graphs (SSGs) and introduces Dynamic Prompt Module (DPM) to guide a Graph Parsing Neural Network (GPNN) in generating action-specific representations. Furthermore, we design a video-adapted GPNN that aggregates information using dynamic weights. Experiments in video action recognition demonstrate the effectiveness of our approach when compared with the state-of-the-art methods. Our code can be found in https://github.com/iamsnaping/ProDA.git

cs.CV↗

arXiv2025

Unfitted Lattice Green's Function Method for Exterior Scattering in Complex Geometry

This paper develops a finite-difference analogue of the boundary integral/element method for the numerical solution of two-dimensional exterior scattering from scatterers of arbitrary shapes. The discrete fundamental solution, known as the lattice Green's function (LGF), for the Helmholtz equation on an infinite lattice is derived and employed to construct boundary algebraic equations through the discrete potentials framework. Unlike the continuous fundamental solution used in boundary integral methods, the LGF introduces no singularity, which simplifies numerical implementation. Boundary conditions are incorporated through local Lagrange interpolation on unfitted cut cells. The resulting method retains key advantages of boundary integral approaches-including dimension reduction and the absence of artificial boundary conditions--while enabling finite differences for complex geometries. Numerical results demonstrate the accuracy and robustness of the method for various scatterers, including circular, triangular, and multiple-body configurations.

math.NA↗

arXiv2025

DW-A-PRM: A Dynamic Weighted Planner

Robot path planning plays a pivotal role in enabling autonomous systems to navigate safely and efficiently in complex and uncertain environments. Despite extensive research on classical graph-based methods and sampling-based planners, achieving an optimal balance between global optimality, computational efficiency, and adaptability to dynamic environments remains an open challenge. To address this issue, this paper proposes a hybrid path planning framework, which integrates heuristic-driven search with probabilistic roadmap construction under a dynamic weighting scheme. By coupling the global guidance of A* with the stochastic exploration of PRM, the method achieves a synergistic balance between search optimality and computational tractability. Comprehensive experiments in diverse simulated environments demonstrate that the proposed method consistently yields smoother and shorter paths while significantly reducing computational overhead compared with conventional approach and other hybrid planners. These results highlight the potential of the proposed framework as an effective and generalizable solution for real-time robotic navigation in complex environments.

cs.RO↗

arXiv2025

LoongRL: Reinforcement Learning for Advanced Reasoning over Long Contexts

Reasoning over long contexts is essential for large language models. While reinforcement learning (RL) enhances short-context reasoning by inducing "Aha" moments in chain-of-thought, the advanced thinking patterns required for long-context reasoning remain largely unexplored, and high-difficulty RL data are scarce. In this paper, we introduce LoongRL, a data-driven RL method for advanced long-context reasoning. Central to LoongRL is KeyChain, a synthesis approach that transforms short multi-hop QA into high-difficulty long-context tasks by inserting UUID chains that hide the true question among large collections of distracting documents. Solving these tasks requires the model to trace the correct chain step-by-step, identify the true question, retrieve relevant facts and reason over them to answer correctly. RL training on KeyChain data induces an emergent plan-retrieve-reason-recheck reasoning pattern that generalizes far beyond training length. Models trained at 16K effectively solve 128K tasks without prohibitive full-length RL rollout costs. On Qwen2.5-7B and 14B, LoongRL substantially improves long-context multi-hop QA accuracy by +23.5% and +21.1% absolute gains. The resulting LoongRL-14B reaches a score of 74.2, rivaling much larger frontier models such as o3-mini (74.5) and DeepSeek-R1 (74.9). It also improves long-context retrieval, passes all 128K needle-in-a-haystack stress tests, and preserves short-context reasoning capabilities.

cs.CL↗

arXiv2025

OneSearch: A Preliminary Exploration of the Unified End-to-End Generative Framework for E-commerce Search

Traditional e-commerce search systems employ multi-stage cascading architectures (MCA) that progressively filter items through recall, pre-ranking, and ranking stages. While effective at balancing computational efficiency with business conversion, these systems suffer from fragmented computation and optimization objective collisions across stages, which ultimately limit their performance ceiling. To address these, we propose \textbf{OneSearch}, the first industrial-deployed end-to-end generative framework for e-commerce search. This framework introduces three key innovations: (1) a Keyword-enhanced Hierarchical Quantization Encoding (KHQE) module, to preserve both hierarchical semantics and distinctive item attributes while maintaining strong query-item relevance constraints; (2) a multi-view user behavior sequence injection strategy that constructs behavior-driven user IDs and incorporates both explicit short-term and implicit long-term sequences to model user preferences comprehensively; and (3) a Preference-Aware Reward System (PARS) featuring multi-stage supervised fine-tuning and adaptive reward-weighted ranking to capture fine-grained user preferences. Extensive offline evaluations on large-scale industry datasets demonstrate OneSearch's superior performance for high-quality recall and ranking. The rigorous online A/B tests confirm its ability to enhance relevance in the same exposure position, achieving statistically significant improvements: +1.67% item CTR, +2.40% buyer, and +3.22% order volume. Furthermore, OneSearch reduces operational expenditure by 75.40% and improves Model FLOPs Utilization from 3.26% to 27.32%. The system has been successfully deployed across multiple search scenarios in Kuaishou, serving millions of users, generating tens of millions of PVs daily.

cs.IR↗

arXiv2025

Nonlinear Symmetry-Fragmentation of Nonabelian Anyons In Symmetry-Enriched Topological Phases: A String-Net Model Realization

Symmetry-enriched topological (SET) phases combine intrinsic topological order with global symmetries, giving rise to novel symmetry phenomena. While SET phases with Abelian anyons are relatively well understood, those involving nonabelian anyons remain elusive. This obscurity stems from the multidimensional internal gauge spaces intrinsic to nonabelian anyons -- a feature first made explicit in [1,2] and further explored and formalized in our recent works [3-8]. These internal spaces can transform in highly nontrivial ways under global symmetries. In this work, we employ an exactly solvable model -- the multifusion Hu-Geer-Wu string-net model introduced in a companion paper [9] -- to reveal how the internal gauge spaces of nonabelian anyons transform under symmetries. We uncover a universal mechanism, global symmetry fragmentation (GSF), whereby symmetry-invariant anyons exhibit internal Hilbert space decompositions into eigensubspaces labeled by generally fractional symmetry charges. Meanwhile, symmetry-permuted anyons hybridize and fragment their internal spaces in accordance with their symmetry behavior. These fragmented structures realize genuinely nonlinear symmetry representations -- to be termed coherent representations -- that transcend conventional linear and projective classifications, reflecting the categorical nature of symmetries in topological phases. Our results identify nonlinear fragmentation as a hallmark of nonabelian SETs and suggest new routes for symmetry-enabled control in topological quantum computation.

cond-mat.str-el↗

arXiv2025

OViP: Online Vision-Language Preference Learning for VLM Hallucination

Large vision-language models (LVLMs) remain vulnerable to hallucination, often generating content misaligned with visual inputs. Although recent training-based approaches aim to mitigate hallucination, they typically rely on predefined or randomly edited negative samples that do not reflect actual model errors, thus limiting training efficacy. In this work, we propose an Online Vision-language Preference Learning (OViP) framework that dynamically constructs contrastive training data based on the model's own hallucinated outputs. By identifying semantic differences between sampled response pairs and synthesizing negative images using a diffusion model, OViP generates more relevant supervision signals in real time. This failure-driven training enables adaptive alignment of both textual and visual preferences. Moreover, we refine existing evaluation protocols to better capture the trade-off between hallucination suppression and expressiveness. Experiments on hallucination and general benchmarks demonstrate that OViP not only reduces hallucinations while preserving core multi-modal capabilities, but also substantially improves training efficiency. Code is available at https://github.com/lsjlsj35/Online-Vision-Language-Preference-Learning-for-VLM-Hallucination.

cs.CV↗

arXiv2025

A framework for rapid, reproducible, and high-fidelity whole-brain multi-pool CEST imaging at 3T

Purpose: To develop and validate a framework for rapid, accurate, and reproducible whole-brain, multi-pool chemical exchange saturation transfer (CEST) imaging at 3T, addressing challenges of long acquisition times and confounding factors. Methods: A single-shot 3D true fast imaging with steady-state precession (True FISP) sequence was optimized for whole-brain multi-pool CEST. Rapid B0, B1, and T1 mapping was performed using a dual-echo modified four-angle method. A feed-forward neural network was developed for rapid B1 correction, trained against the conventional multi-power method. The apparent exchange-dependent relaxation (AREX) metric was used to correct for T1 and magnetization transfer (MT) effects. The framework was validated in phantoms and healthy human subjects (N=8), including a test-retest reproducibility assessment. Results: The True FISP sequence yielded high-quality, whole-brain images with minimal artifacts and distortion in a clinically feasible scan time (~9 minutes). Phantom studies confirmed the effectiveness of B1 correction (coefficient of variation [CV] for MT_MTRLD decreased from 22.49% to 4.61%) and AREX-based confounder correction (CV for APT_AREX reduced from 33.6% to 6.9%). The neural network B1 correction showed excellent agreement with the conventional multi-power method in vivo (ICC > 0.97). High test-retest reproducibility was demonstrated across 96 brain regions, with the average CV for APT_AREX under 10% for over 95% of regions. Conclusion: A rapid and robust framework for whole-brain quantitative multi-pool CEST imaging was successfully developed and validated. By integrating an efficient acquisition sequence with a streamlined correction pipeline, this approach overcomes key barriers to clinical translation, enabling reliable metabolic imaging for widespread brain pathologies.

physics.med-ph↗

arXiv2025

Diagnosing Memorization in Chain-of-Thought Reasoning, One Token at a Time

Large Language Models (LLMs) perform well on reasoning benchmarks but often fail when inputs alter slightly, raising concerns about the extent to which their success relies on memorization. This issue is especially acute in Chain-of-Thought (CoT) reasoning, where spurious memorized patterns can trigger intermediate errors that cascade into incorrect final answers. We introduce STIM, a novel framework for Source-aware Token-level Identification of Memorization, which attributes each token in a reasoning chain to one of multiple memorization sources - local, mid-range, or long-range - based on their statistical co-occurrence with the token in the pretraining corpus. Our token-level analysis across tasks and distributional settings reveals that models rely more on memorization in complex or long-tail cases, and that local memorization is often the dominant driver of errors, leading to up to 67% of wrong tokens. We also show that memorization scores from STIM can be effective in predicting the wrong tokens in the wrong reasoning step. STIM offers a powerful tool for diagnosing and improving model reasoning and can generalize to other structured step-wise generation tasks.

cs.CL↗

arXiv2025

Simple o3: Towards Interleaved Vision-Language Reasoning

Multimodal Large Language Models (MLLMs) have shown impressive performance on vision-language tasks, but their long Chain-of-Thought (CoT) capabilities in multimodal scenarios remain underexplored. Inspired by OpenAI's o3 model, which emulates human-like ''thinking with image'' through iterative visual transformations and linguistic reasoning, we propose Simple o3, an end-to-end framework that integrates dynamic tool interactions (e.g., cropping, zooming, and reusing) into interleaved vision-language reasoning via supervised fine-tuning (SFT). Our approach features a scalable data synthesis pipeline that generates high-quality interleaved vision-language reasoning chains via an ''observe-reason-act'' cycle, complete with executable visual operations and rigorous verification, yielding the open-source TWI-Tools-146K dataset. Experimental results demonstrate Simple o3's superior performance on diverse benchmarks, outperforming existing approaches. By combining enhanced reasoning capabilities, Simple o3 establishes a powerful yet computationally affordable paradigm for advancing multimodal reasoning. Remarkably, we provide the first in-depth analysis of different interleaved reasoning strategies, offering insights into their impact on model performance. We found that by introducing additional visual tokens for interleaved vision-language reasoning, reusing and magnifying the original image significantly improves the model's visual reasoning and fine-grained perception, while image cropping based on precise visual grounding allows the model to effectively focus on key entities or regions, further enhancing its capabilities.

cs.CV↗

arXiv2025

Marine Chlorophyll Prediction and Driver Analysis based on LSTM-RF Hybrid Models

Marine chlorophyll concentration is an important indicator of ecosystem health and carbon cycle strength, and its accurate prediction is crucial for red tide warning and ecological response. In this paper, we propose a LSTM-RF hybrid model that combines the advantages of LSTM and RF, which solves the deficiencies of a single model in time-series modelling and nonlinear feature portrayal. Trained with multi-source ocean data(temperature, salinity, dissolved oxygen, etc.), the experimental results show that the LSTM-RF model has an R^2 of 0.5386, an MSE of 0.005806, and an MAE of 0.057147 on the test set, which is significantly better than using LSTM (R^2 = 0.0208) and RF (R^2 =0.4934) alone , respectively. The standardised treatment and sliding window approach improved the prediction accuracy of the model and provided an innovative solution for high-frequency prediction of marine ecological variables.

cs.LG↗

arXiv2025

Physics-Informed Recurrent Network for State-Space Modeling of Gas Pipeline Networks

As a part of the integrated energy system (IES), gas pipeline networks can provide additional flexibility to power systems through coordinated optimal dispatch. An accurate pipeline network model is critical for the optimal operation and control of IESs. However, inaccuracies or unavailability of accurate pipeline parameters often introduce errors in the state-space models of such networks. This paper proposes a physics-informed recurrent network (PIRN) to identify the state-space model of gas pipelines. It fuses sparse measurement data with fluid-dynamic behavior expressed by partial differential equations. By embedding the physical state-space model within the recurrent network, parameter identification becomes an end-to-end PIRN training task. The model can be realized in PyTorch through modifications to a standard RNN backbone. Case studies demonstrate that our proposed PIRN can accurately estimate gas pipeline models from sparse terminal node measurements, providing robust performance and significantly higher parameter efficiency. Furthermore, the identified state-space model of the pipeline network can be seamlessly integrated into optimization frameworks.

eess.SY↗

arXiv2025

Non-Iterative Coordination of Interconnected Power Grids via Dimension-Decomposition-Based Flexibility Aggregation

The bulk power grid is divided into regional grids interconnected with multiple tie-lines for efficient operation. Since interconnected power grids are operated by different control centers, it is a challenging task to realize coordinated dispatch of multiple regional grids. A viable solution is to compute a flexibility aggregation model for each regional power grid, then optimize the tie-line schedule using the aggregated models to implement non-iterative coordinated dispatch. However, challenges such as intricate interdependencies and curse of dimensionality persist in computing the aggregated models in high-dimensional space. Existing methods like Fourier-Motzkin elimination, vertex search, and multi-parameter programming are limited by dimensionality and conservatism, hindering practical application. This paper presents a novel dimension-decomposition-based flexibility aggregation algorithm for calculating the aggregated models of multiple regional power grids, enabling non-iterative coordination in large-scale interconnected systems. Compared to existing methods, the proposed approach yields a significantly less conservative flexibility region. The derived flexibility aggregation model for each regional power grid has a well-defined physical counterpart, which facilitates intuitive analysis of multi-port regional power grids and provides valuable insights into their internal resource endowments. Numerical tests validate the feasibility of the aggregated model and demonstrate its accuracy in coordinating interconnected power grids.

eess.SY↗

arXiv2025

OneSug: The Unified End-to-End Generative Framework for E-commerce Query Suggestion

Query suggestion plays a crucial role in enhancing user experience in e-commerce search systems by providing relevant query recommendations that align with users' initial input. This module helps users navigate towards personalized preference needs and reduces typing effort, thereby improving search experience. Traditional query suggestion modules usually adopt multi-stage cascading architectures, for making a well trade-off between system response time and business conversion. But they often suffer from inefficiencies and suboptimal performance due to inconsistent optimization objectives across stages. To address these, we propose OneSug, the first end-to-end generative framework for e-commerce query suggestion. OneSug incorporates a prefix2query representation enhancement module to enrich prefixes using semantically and interactively related queries to bridge content and business characteristics, an encoder-decoder generative model that unifies the query suggestion process, and a reward-weighted ranking strategy with behavior-level weights to capture fine-grained user preferences. Extensive evaluations on large-scale industry datasets demonstrate OneSug's ability for effective and efficient query suggestion. Furthermore, OneSug has been successfully deployed for the entire traffic on the e-commerce search engine in Kuaishou platform for over 1 month, with statistically significant improvements in user top click position (-9.33%), CTR (+2.01%), Order (+2.04%), and Revenue (+1.69%) over the online multi-stage strategy, showing great potential in e-commercial conversion.

cs.IR↗

arXiv2025

AutoJudger: An Agent-Driven Framework for Efficient Benchmarking of MLLMs

Evaluating multimodal large language models (MLLMs) is increasingly expensive, as the growing size and cross-modality complexity of benchmarks demand significant scoring efforts. To tackle with this difficulty, we introduce AutoJudger, an agent-driven framework for efficient and adaptive benchmarking of MLLMs that tackles this escalating cost. AutoJudger employs the Item Response Theory (IRT) to estimate the question difficulty and an autonomous evaluation agent to dynamically select the most informative test questions based on the model's real-time performance. Specifically, AutoJudger incorporates two pivotal components: a semantic-aware retrieval mechanism to ensure that selected questions cover diverse and challenging scenarios across both vision and language modalities, and a dynamic memory that maintains contextual statistics of previously evaluated questions to guide coherent and globally informed question selection throughout the evaluation process. Extensive experiments on four representative multimodal benchmarks demonstrate that our adaptive framework dramatically reduces evaluation expenses, i.e. AutoJudger uses only 4% of the data to achieve over 90% ranking accuracy with the full benchmark evaluation on MMT-Bench.

cs.CL↗

arXiv2025

Classification of Gapped Domain Walls of Topological Orders in 2+1 dimensions: A Levin-Wen Model Realization

This paper introduces a novel systematic construction of gapped domain walls (GDWs) within the Levin-Wen (LW) model. By gluing two LW models along their open sides in a compatible way, we achieve a complete GDW classification by subsets of bulk input data, which encompass the classifications in terms of bimodule categories. A generalized bimodule structure is introduced to capture domain-wall excitations. Furthermore, we demonstrate that folding along any GDW yields a gapped boundary (GB) described by a Frobenius algebra of the input UFC for the folded model, thus bridging our GDW classification and the GB classification in \cite{hu2018boundary}.

cond-mat.str-el↗

arXiv2025

VGDFR: Diffusion-based Video Generation with Dynamic Latent Frame Rate

Diffusion Transformer(DiT)-based generation models have achieved remarkable success in video generation. However, their inherent computational demands pose significant efficiency challenges. In this paper, we exploit the inherent temporal non-uniformity of real-world videos and observe that videos exhibit dynamic information density, with high-motion segments demanding greater detail preservation than static scenes. Inspired by this temporal non-uniformity, we propose VGDFR, a training-free approach for Diffusion-based Video Generation with Dynamic Latent Frame Rate. VGDFR adaptively adjusts the number of elements in latent space based on the motion frequency of the latent space content, using fewer tokens for low-frequency segments while preserving detail in high-frequency segments. Specifically, our key contributions are: (1) A dynamic frame rate scheduler for DiT video generation that adaptively assigns frame rates for video segments. (2) A novel latent-space frame merging method to align latent representations with their denoised counterparts before merging those redundant in low-resolution space. (3) A preference analysis of Rotary Positional Embeddings (RoPE) across DiT layers, informing a tailored RoPE strategy optimized for semantic and local information capture. Experiments show that VGDFR can achieve a speedup up to 3x for video generation with minimal quality degradation.

cs.CV↗