arXiv subjects

Fan Wu

Publications and source records attributed to Fan Wu.

At least 181 records · Page 10Linked to original sources

arXiv2025

Collaborative Learning of On-Device Small Model and Cloud-Based Large Model: Advances and Future Directions

The conventional cloud-based large model learning framework is increasingly constrained by latency, cost, personalization, and privacy concerns. In this survey, we explore an emerging paradigm: collaborative learning between on-device small model and cloud-based large model, which promises low-latency, cost-efficient, and personalized intelligent services while preserving user privacy. We provide a comprehensive review across hardware, system, algorithm, and application layers. At each layer, we summarize key problems and recent advances from both academia and industry. In particular, we categorize collaboration algorithms into data-based, feature-based, and parameter-based frameworks. We also review publicly available datasets and evaluation metrics with user-level or device-level consideration tailored to collaborative learning settings. We further highlight real-world deployments, ranging from recommender systems and mobile livestreaming to personal intelligent assistants. We finally point out open research directions to guide future development in this rapidly evolving field.

cs.LG↗

arXiv2025

Efficient Distributed Retrieval-Augmented Generation for Enhancing Language Model Performance

Small language models (SLMs) support efficient deployments on resource-constrained edge devices, but their limited capacity compromises inference performance. Retrieval-augmented generation (RAG) is a promising solution to enhance model performance by integrating external databases, without requiring intensive on-device model retraining. However, large-scale public databases and user-specific private contextual documents are typically located on the cloud and the device separately, while existing RAG implementations are primarily centralized. To bridge this gap, we propose DRAGON, a distributed RAG framework to enhance on-device SLMs through both general and personal knowledge without the risk of leaking document privacy. Specifically, DRAGON decomposes multi-document RAG into multiple parallel token generation processes performed independently and locally on the cloud and the device, and employs a newly designed Speculative Aggregation, a dual-side speculative algorithm to avoid frequent output synchronization between the cloud and device. A new scheduling algorithm is further introduced to identify the optimal aggregation side based on real-time network conditions. Evaluations on real-world hardware testbed demonstrate a significant performance improvement of DRAGON-up to 1.9x greater gains over standalone SLM compared to the centralized RAG, substantial reduction in per-token latency, and negligible Time to First Token (TTFT) overhead.

cs.LG↗

arXiv2025

Double-optical phase-transition in a three level Rydberg state in thermal Rubidium vapor

We report on the observation of electromagnetically induced transparency (EIT) with intrinsic phase transitions in a three-level ladder system within rubidium atomic vapor. The observed abrupt transitions between low and high Rydberg occupancy states manifest in the probe beam transmission, depending on the principal quantum number, the Rabi frequency of the coupling field, atomic density, and probe beam detuning. Our study elucidates the underlying interaction mechanisms governing the EIT phase transition and enriches the existing experiments of multi-parameter regulation phase transitions. These findings establish a robust platform for investigating nonequilibrium phase transitions in atomic ensembles, bridging the gap between classical mean-field theories and microscopic quantum dynamics.

physics.optics↗

arXiv2025

Embedding groups into boundedly acyclic groups

We show that the \sϕ-labeled Thompson groups and the twisted Brin--Thompson groups are boundedly acyclic. This allows us to prove several new embedding results for groups. First, every group of type $F_n$ embeds quasi-isometrically into a boundedly acyclic group of type $F_n$ that has no proper finite index subgroups. This improves a result of Bridson and a theorem of Fournier-Facio--Löh--Moraschini. Second, every group of type $F_n$ embeds quasi-isometrically into a $5$-uniformly perfect group of type $F_n$. Third, using Belk--Zaremsky's construction of twisted Brin--Thompson groups, we show that every finitely generated group embeds quasi-isometrically into a finitely generated boundedly acyclic simple group. We also partially answer some questions of Brothier and Tanushevski regarding the finiteness property of $ϕ$-labeled Thompson group $V_ϕ(G)$ and $F_ϕ(G)$.

math.GR↗

arXiv2025

FeatInsight: An Online ML Feature Management System on 4Paradigm Sage-Studio Platform

Feature management is essential for many online machine learning applications and can often become the performance bottleneck (e.g., taking up to 70% of the overall latency in sales prediction service). Improper feature configurations (e.g., introducing too many irrelevant features) can severely undermine the model's generalization capabilities. However, managing online ML features is challenging due to (1) large-scale, complex raw data (e.g., the 2018 PHM dataset contains 17 tables and dozens to hundreds of columns), (2) the need for high-performance, consistent computation of interdependent features with complex patterns, and (3) the requirement for rapid updates and deployments to accommodate real-time data changes. In this demo, we present FeatInsight, a system that supports the entire feature lifecycle, including feature design, storage, visualization, computation, verification, and lineage management. FeatInsight (with OpenMLDB as the execution engine) has been deployed in over 100 real-world scenarios on 4Paradigm's Sage Studio platform, handling up to a trillion-dimensional feature space and enabling millisecond-level feature updates. We demonstrate how FeatInsight enhances feature design efficiency (e.g., for online product recommendation) and improve feature computation performance (e.g., for online fraud detection). The code is available at https://github.com/4paradigm/FeatInsight.

cs.DB↗

arXiv2025

Efficient Twin Migration in Vehicular Metaverses: Multi-Agent Split Deep Reinforcement Learning with Spatio-Temporal Trajectory Generation

Vehicle Twins (VTs) as digital representations of vehicles can provide users with immersive experiences in vehicular metaverse applications, e.g., Augmented Reality (AR) navigation and embodied intelligence. VT migration is an effective way that migrates the VT when the locations of physical entities keep changing to maintain seamless immersive VT services. However, an efficient VT migration is challenging due to the rapid movement of vehicles, dynamic workloads of Roadside Units (RSUs), and heterogeneous resources of the RSUs. To achieve efficient migration decisions and a minimum latency for the VT migration, we propose a multi-agent split Deep Reinforcement Learning (DRL) framework combined with spatio-temporal trajectory generation. In this framework, multiple split DRL agents utilize split architecture to efficiently determine VT migration decisions. Furthermore, we propose a spatio-temporal trajectory generation algorithm based on trajectory datasets and road network data to simulate vehicle trajectories, enhancing the generalization of the proposed scheme for managing VT migration in dynamic network environments. Finally, experimental results demonstrate that the proposed scheme not only enhances the Quality of Experience (QoE) by 29% but also reduces the computational parameter count by approximately 25% while maintaining similar performances, enhancing users' immersive experiences in vehicular metaverses.

cs.NI↗

arXiv2025

OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotations

Document content extraction is a critical task in computer vision, underpinning the data needs of large language models (LLMs) and retrieval-augmented generation (RAG) systems. Despite recent progress, current document parsing methods have not been fairly and comprehensively evaluated due to the narrow coverage of document types and the simplified, unrealistic evaluation procedures in existing benchmarks. To address these gaps, we introduce OmniDocBench, a novel benchmark featuring high-quality annotations across nine document sources, including academic papers, textbooks, and more challenging cases such as handwritten notes and densely typeset newspapers. OmniDocBench supports flexible, multi-level evaluations--ranging from an end-to-end assessment to the task-specific and attribute--based analysis using 19 layout categories and 15 attribute labels. We conduct a thorough evaluation of both pipeline-based methods and end-to-end vision-language models, revealing their strengths and weaknesses across different document types. OmniDocBench sets a new standard for the fair, diverse, and fine-grained evaluation in document parsing. Dataset and code are available at https://github.com/opendatalab/OmniDocBench.

cs.CV↗

arXiv2025

Image Over Text: Transforming Formula Recognition Evaluation with Character Detection Matching

Formula recognition presents significant challenges due to the complicated structure and varied notation of mathematical expressions. Despite continuous advancements in formula recognition models, the evaluation metrics employed by these models, such as BLEU and Edit Distance, still exhibit notable limitations. They overlook the fact that the same formula has diverse representations and is highly sensitive to the distribution of training data, thereby causing unfairness in formula recognition evaluation. To this end, we propose a Character Detection Matching (CDM) metric, ensuring the evaluation objectivity by designing an image-level rather than a LaTeX-level metric score. Specifically, CDM renders both the model-predicted LaTeX and the ground-truth LaTeX formulas into image-formatted formulas, then employs visual feature extraction and localization techniques for precise character-level matching, incorporating spatial position information. Such a spatially-aware and character-matching method offers a more accurate and equitable evaluation compared with previous BLEU and Edit Distance metrics that rely solely on text-based character matching. Experimentally, we evaluated various formula recognition models using CDM, BLEU, and ExpRate metrics. Their results demonstrate that the CDM aligns more closely with human evaluation standards and provides a fairer comparison across different models by eliminating discrepancies caused by diverse formula representations. Code is available at https://github.com/opendatalab/UniMERNet/tree/main/cdm.

cs.CV↗

arXiv2025

Switching on and off the spin polarization of the conduction band in antiferromagnetic bilayer transistors

Antiferromagnetic conductors with suitably broken spatial symmetries host spin-polarized bands, which lead to transport phenomena commonly observed in metallic ferromagnets. In bulk materials, it is the given crystalline structure that determines whether symmetries are broken and spin-polarized bands are present. Here we demonstrate experimentally that double-gate transistors realized on bilayers of van der Waals antiferromagnetic semiconductor CrPS4 allow the relevant symmetry to be controlled by a perpendicular electric displacement field. Such a level of control enables the spin-polarization of the conduction band to be switched on and off. Because conduction band states with opposite spin-polarizations are hosted in the different layers and are spatially separated, these devices also give control over the magnetization of the electrons that are accumulated electrostatically. Our experiments show that double-gated CrPS4 transistors provide a viable platform to create gate-induced conductors with near unity spin polarization at the Fermi level, as well as devices with a full electrostatic control of the total magnetization of the system.

cond-mat.mes-hall↗

arXiv2025

CFNet: Optimizing Remote Sensing Change Detection through Content-Aware Enhancement

Change detection is a crucial and widely applied task in remote sensing, aimed at identifying and analyzing changes occurring in the same geographical area over time. Due to variability in acquisition conditions, bi-temporal remote sensing images often exhibit significant differences in image style. Even with the powerful generalization capabilities of DNNs, these unpredictable style variations between bi-temporal images inevitably affect model's ability to accurately detect changed areas. To address issue above, we propose the Content Focuser Network (CFNet), which takes content-aware strategy as a key insight. CFNet employs EfficientNet-B5 as the backbone for feature extraction. To enhance the model's focus on the content features of images while mitigating the misleading effects of style features, we develop a constraint strategy that prioritizes the content features of bi-temporal images, termed Content-Aware. Furthermore, to enable the model to flexibly focus on changed and unchanged areas according to the requirements of different stages, we design a reweighting module based on the cosine distance between bi-temporal image features, termed Focuser. CFNet achieve outstanding performance across three well-known change detection datasets: CLCD (F1: 81.41%, IoU: 68.65%), LEVIR-CD (F1: 92.18%, IoU: 85.49%), and SYSU-CD (F1: 82.89%, IoU: 70.78%). The code and pretrained models of CFNet are publicly released at https://github.com/wifiBlack/CFNet.

cs.CV↗

arXiv2025

LEMMo-Plan: LLM-Enhanced Learning from Multi-Modal Demonstration for Planning Sequential Contact-Rich Manipulation Tasks

Large Language Models (LLMs) have gained popularity in task planning for long-horizon manipulation tasks. To enhance the validity of LLM-generated plans, visual demonstrations and online videos have been widely employed to guide the planning process. However, for manipulation tasks involving subtle movements but rich contact interactions, visual perception alone may be insufficient for the LLM to fully interpret the demonstration. Additionally, visual data provides limited information on force-related parameters and conditions, which are crucial for effective execution on real robots. In this paper, we introduce an in-context learning framework that incorporates tactile and force-torque information from human demonstrations to enhance LLMs' ability to generate plans for new task scenarios. We propose a bootstrapped reasoning pipeline that sequentially integrates each modality into a comprehensive task plan. This task plan is then used as a reference for planning in new task configurations. Real-world experiments on two different sequential manipulation tasks demonstrate the effectiveness of our framework in improving LLMs' understanding of multi-modal demonstrations and enhancing the overall planning performance.

cs.RO↗

arXiv2025

Policy Constraint by Only Support Constraint for Offline Reinforcement Learning

Offline reinforcement learning (RL) aims to optimize a policy by using pre-collected datasets, to maximize cumulative rewards. However, offline reinforcement learning suffers challenges due to the distributional shift between the learned and behavior policies, leading to errors when computing Q-values for out-of-distribution (OOD) actions. To mitigate this issue, policy constraint methods aim to constrain the learned policy's distribution with the distribution of the behavior policy or confine action selection within the support of the behavior policy. However, current policy constraint methods tend to exhibit excessive conservatism, hindering the policy from further surpassing the behavior policy's performance. In this work, we present Only Support Constraint (OSC) which is derived from maximizing the total probability of learned policy in the support of behavior policy, to address the conservatism of policy constraint. OSC presents a regularization term that only restricts policies to the support without imposing extra constraints on actions within the support. Additionally, to fully harness the performance of the new policy constraints, OSC utilizes a diffusion model to effectively characterize the support of behavior policies. Experimental evaluations across a variety of offline RL benchmarks demonstrate that OSC significantly enhances performance, alleviating the challenges associated with distributional shifts and mitigating conservatism of policy constraints. Code is available at https://github.com/MoreanP/OSC.

cs.LG↗

arXiv2025

TacDiffusion: Force-domain Diffusion Policy for Precise Tactile Manipulation

Assembly is a crucial skill for robots in both modern manufacturing and service robotics. However, mastering transferable insertion skills that can handle a variety of high-precision assembly tasks remains a significant challenge. This paper presents a novel framework that utilizes diffusion models to generate 6D wrench for high-precision tactile robotic insertion tasks. It learns from demonstrations performed on a single task and achieves a zero-shot transfer success rate of 95.7% across various novel high-precision tasks. Our method effectively inherits the self-adaptability demonstrated by our previous work. In this framework, we address the frequency misalignment between the diffusion policy and the real-time control loop with a dynamic system-based filter, significantly improving the task success rate by 9.15%. Furthermore, we provide a practical guideline regarding the trade-off between diffusion models' inference ability and speed.

cs.RO↗

arXiv2025

It's My Data Too: Private ML for Datasets with Multi-User Training Examples

We initiate a study of algorithms for model training with user-level differential privacy (DP), where each example may be attributed to multiple users, which we call the multi-attribution model. We first provide a carefully chosen definition of user-level DP under the multi-attribution model. Training in the multi-attribution model is facilitated by solving the contribution bounding problem, i.e. the problem of selecting a subset of the dataset for which each user is associated with a limited number of examples. We propose a greedy baseline algorithm for the contribution bounding problem. We then empirically study this algorithm for a synthetic logistic regression task and a transformer training task, including studying variants of this baseline algorithm that optimize the subset chosen using different techniques and criteria. We find that the baseline algorithm remains competitive with its variants in most settings, and build a better understanding of the practical importance of a bias-variance tradeoff inherent in solutions to the contribution bounding problem.

cs.LG↗

arXiv2025

Modulation of supernarrow EIT pair via atomic coherence

We report the phenomena of electromagnetically induced transparency (EIT) and electromagnetically induced absorption (EIA) using two identical beams in rubidium atomic vapor. The Λ-type EIT configuration is employed to examine the EIT spectrum for the D1 line in 87Rb F=2 characteristics16 by varying parameters such as frequency detuning, Iprobe/Ipump, the total power of probe and pump beam. Notably, the pump beam is also investigated in this process, which has not been previously studied. We study the effect of of the phase between the two applied fields and find that EIA and EIT can transform into each other by adjusting the relative phase. These finding may have applications in light drag or storage, optical switching, and sensing.

physics.atom-ph↗

arXiv2025

Online Pseudo-average Shifting Attention(PASA) for Robust Low-precision LLM Inference: Algorithms and Numerical Analysis

Attention calculation is extremely time-consuming for long-sequence inference tasks, such as text or image/video generation, in large models. To accelerate this process, we developed a low-precision, mathematically-equivalent algorithm called PASA, based on Flash Attention. PASA introduces two novel techniques: online pseudo-average shifting and global recovering. These techniques enable the use of half-precision computation throughout the Flash Attention process without incurring overflow instability or unacceptable numerical accuracy loss. This algorithm enhances performance on memory-restricted AI hardware architectures, such as the Ascend Neural-network Processing Unit(NPU), by reducing data movement and increasing computational FLOPs. The algorithm is validated using both designed random benchmarks and real large models. We find that the large bias and amplitude of attention input data are critical factors contributing to numerical overflow ($>65504$ for half precision) in two different categories of large models (Qwen2-7B language models and Stable-Video-Diffusion multi-modal models). Specifically, overflow arises due to the large bias in the sequence dimension and the resonance mechanism between the query and key in the head dimension of the Stable-Video-Diffusion models. The resonance mechanism is defined as phase coincidence or 180-degree phase shift between query and key matrices. It will remarkably amplify the element values of attention score matrix. This issue also applies to the Qwen models. Additionally, numerical accuracy is assessed through root mean square error (RMSE) and by comparing the final generated texts and videos to those produced using high-precision attention.

cs.LG↗

arXiv2025

Efficient cavity-mediated energy transfer between photosynthetic light harvesting complexes from strong to weak coupling regime

Excitation energy transfer between photosynthetic light-harvesting complexes is vital for highly efficient primary photosynthesis. Controlling this process is the key for advancing the emerging artificial photosynthetic systems. Here, we experimentally demonstrate the enhanced excitation energy transfer between photosynthetic light-harvesting 2 complexes (LH2) mediated through the Fabry-Perot optical microcavity. Using intensity-dependent pump-probe spectroscopy, we analyse the exciton-exciton annihilation (EEA) due to inter-LH2 energy transfer. Comparing EEA in LH2 within cavity samples and the bare LH2 films, we observe enhanced EEA in cavities indicating improved excitation energy transfer via coupling to a common cavity mode. Surprisingly, the effect remains even in the weak coupling regime. The enhancement is attributed to the additional connectivity between LH2s introduced by the resonant optical microcavity. Our results suggest that optical microcavities can be a strategic tool for modifying excitation energy transfer between molecular complexes, offering a promising approach towards efficient artificial light harvesting.

physics.optics↗

arXiv2025

Ten Challenging Problems in Federated Foundation Models

Federated Foundation Models (FedFMs) represent a distributed learning paradigm that fuses general competences of foundation models as well as privacy-preserving capabilities of federated learning. This combination allows the large foundation models and the small local domain models at the remote clients to learn from each other in a teacher-student learning setting. This paper provides a comprehensive summary of the ten challenging problems inherent in FedFMs, encompassing foundational theory, utilization of private data, continual learning, unlearning, Non-IID and graph data, bidirectional knowledge transfer, incentive mechanism design, game mechanism design, model watermarking, and efficiency. The ten challenging problems manifest in five pivotal aspects: ``Foundational Theory," which aims to establish a coherent and unifying theoretical framework for FedFMs. ``Data," addressing the difficulties in leveraging domain-specific knowledge from private data while maintaining privacy; ``Heterogeneity," examining variations in data, model, and computational resources across clients; ``Security and Privacy," focusing on defenses against malicious attacks and model theft; and ``Efficiency," highlighting the need for improvements in training, communication, and parameter efficiency. For each problem, we offer a clear mathematical definition on the objective function, analyze existing methods, and discuss the key challenges and potential solutions. This in-depth exploration aims to advance the theoretical foundations of FedFMs, guide practical implementations, and inspire future research to overcome these obstacles, thereby enabling the robust, efficient, and privacy-preserving FedFMs in various real-world applications.

cs.LG↗