arXiv subjects

Fan Zhang

Publications and source records attributed to Fan Zhang.

At least 379 records · Page 21Linked to original sources

arXiv2025

FourierFlow: Frequency-aware Flow Matching for Generative Turbulence Modeling

Modeling complex fluid systems, especially turbulence governed by partial differential equations (PDEs), remains a fundamental challenge in science and engineering. Recently, diffusion-based generative models have gained attention as a powerful approach for these tasks, owing to their capacity to capture long-range dependencies and recover hierarchical structures. However, we present both empirical and theoretical evidence showing that generative models struggle with significant spectral bias and common-mode noise when generating high-fidelity turbulent flows. Here we propose FourierFlow, a novel generative turbulence modeling framework that enhances the frequency-aware learning by both implicitly and explicitly mitigating spectral bias and common-mode noise. FourierFlow comprises three key innovations. Firstly, we adopt a dual-branch backbone architecture, consisting of a salient flow attention branch with local-global awareness to focus on sensitive turbulence areas. Secondly, we introduce a frequency-guided Fourier mixing branch, which is integrated via an adaptive fusion strategy to explicitly mitigate spectral bias in the generative model. Thirdly, we leverage the high-frequency modeling capabilities of the masked auto-encoder pre-training and implicitly align the features of the generative model toward high-frequency components. We validate the effectiveness of FourierFlow on three canonical turbulent flow scenarios, demonstrating superior performance compared to state-of-the-art methods. Furthermore, we show that our model exhibits strong generalization capabilities in challenging settings such as out-of-distribution domains, long-term temporal extrapolation, and robustness to noisy inputs. The code can be found at https://github.com/AI4Science-WestlakeU/FourierFlow.

cs.LG↗

arXiv2025

TRiMM: Transformer-Based Rich Motion Matching for Real-Time multi-modal Interaction in Digital Humans

Large Language Model (LLM)-driven digital humans have sparked a series of recent studies on co-speech gesture generation systems. However, existing approaches struggle with real-time synthesis and long-text comprehension. This paper introduces Transformer-Based Rich Motion Matching (TRiMM), a novel multi-modal framework for real-time 3D gesture generation. Our method incorporates three modules: 1) a cross-modal attention mechanism to achieve precise temporal alignment between speech and gestures; 2) a long-context autoregressive model with a sliding window mechanism for effective sequence modeling; 3) a large-scale gesture matching system that constructs an atomic action library and enables real-time retrieval. Additionally, we develop a lightweight pipeline implemented in the Unreal Engine for experimentation. Our approach achieves real-time inference at 120 fps and maintains a per-sentence latency of 0.15 seconds on consumer-grade GPUs (Geforce RTX3060). Extensive subjective and objective evaluations on the ZEGGS, and BEAT datasets demonstrate that our model outperforms current state-of-the-art methods. TRiMM enhances the speed of co-speech gesture generation while ensuring gesture quality, enabling LLM-driven digital humans to respond to speech in real time and synthesize corresponding gestures. Our code is available at https://github.com/teroon/TRiMM-Transformer-Based-Rich-Motion-Matching

cs.GR↗

arXiv2025

A Constructed Response: Designing and Choreographing Robot Arm Movements in Collaborative Dance Improvisation

Dancers often prototype movements themselves or with each other during improvisation and choreography. How are these interactions altered when physically manipulable technologies are introduced into the creative process? To understand how dancers design and improvise movements while working with instruments capable of non-humanoid movements, we engaged dancers in workshops to co-create movements with a robot arm in one-human-to-one-robot and three-human-to-one-robot settings. We found that dancers produced more fluid movements in one-to-one scenarios, experiencing a stronger sense of connection and presence with the robot as a co-dancer. In three-to-one scenarios, the dancers divided their attention between the human dancers and the robot, resulting in increased perceived use of space and more stop-and-go movements, perceiving the robot as part of the stage background. This work highlights how technologies can drive creativity in movement artists adapting to new ways of working with physical instruments, contributing design insights supporting artistic collaborations with non-humanoid agents.

cs.RO↗

arXiv2025

Offline Map Matching Based on Localization Error Distribution Modeling

Offline map matching involves aligning historical trajectories of mobile objects, which may have positional errors, with digital maps. This is essential for applications in intelligent transportation systems (ITS), such as route analysis and traffic pattern mining. Existing methods have two main limitations: (i) they assume a uniform Localization Error Distribution (LED) across urban areas, neglecting environmental factors that lead to suboptimal path search ranges, and (ii) they struggle to efficiently handle local non-shortest paths and detours. To address these issues, we propose a novel offline map matching method for sparse trajectories, called LNSP, which integrates LED modeling and non-shortest path detection. Key innovations include: (i) leveraging public transit trajectories with fixed routes to model LED in finer detail across different city regions, optimizing path search ranges, and (ii) scoring paths using sub-region dependency LED and a sliding window, which reduces global map matching errors. Experimental results using real-world bus and taxi trajectory datasets demonstrate that the LNSP algorithm significantly outperforms existing methods in both efficiency and matching accuracy.

cs.SI↗

arXiv2025

Quantum phase transition in a 2D atomic Bose gas with a g-wave Feshbach resonance

Recent repoet on the formation of two-dimensional Bose-Einstein condensates (BECs) of spinning g-wave molecules is surprise. Here we study quantum phase transition in the quasi-2D atomic Bose gas with a g-wave Feshbach resonance, and show that there are two phase transitions in this system: from a phase with only a atomic Bose-Einstein condensate to a phase with coexistence of atomic and molecular condensation and from the coexistence phase to a phase with only a molecular Bose-Einstein condensate. We show that the g-wave resonance Feshbach has two features in the phase transitions, a phase with only a atomic Bose-Einstein condensate is no longer forbidden and resonance effects appear only at levels beyond the mean-field. We determine the T=0 beyond mean-field phase diagram of the gas as a function of magnetic field and molecular condensation density. We also detemine the ratio of atomic to molecular condensation in the coexistence phase, which can be tested in experiments.

quant-ph↗

arXiv2025

End-to-End Vision Tokenizer Tuning

Existing vision tokenization isolates the optimization of vision tokenizers from downstream training, implicitly assuming the visual tokens can generalize well across various tasks, e.g., image generation and visual question answering. The vision tokenizer optimized for low-level reconstruction is agnostic to downstream tasks requiring varied representations and semantics. This decoupled paradigm introduces a critical misalignment: The loss of the vision tokenization can be the representation bottleneck for target tasks. For example, errors in tokenizing text in a given image lead to poor results when recognizing or generating them. To address this, we propose ETT, an end-to-end vision tokenizer tuning approach that enables joint optimization between vision tokenization and target autoregressive tasks. Unlike prior autoregressive models that use only discrete indices from a frozen vision tokenizer, ETT leverages the visual embeddings of the tokenizer codebook, and optimizes the vision tokenizers end-to-end with both reconstruction and caption objectives. ETT can be seamlessly integrated into existing training pipelines with minimal architecture modifications. Our ETT is simple to implement and integrate, without the need to adjust the original codebooks or architectures of the employed large language models. Extensive experiments demonstrate that our proposed end-to-end vision tokenizer tuning unlocks significant performance gains, i.e., 2-6% for multimodal understanding and visual generation tasks compared to frozen tokenizer baselines, while preserving the original reconstruction capability. We hope this very simple and strong method can empower multimodal foundation models besides image generation and understanding.

cs.CV↗

arXiv2025

GradStop: Exploring Training Dynamics in Unsupervised Outlier Detection through Gradient

Unsupervised Outlier Detection (UOD) is a critical task in data mining and machine learning, aiming to identify instances that significantly deviate from the majority. Without any label, deep UOD methods struggle with the misalignment between the model's direct optimization goal and the final performance goal of Outlier Detection (OD) task. Through the perspective of training dynamics, this paper proposes an early stopping algorithm to optimize the training of deep UOD models, ensuring they perform optimally in OD rather than overfitting the entire contaminated dataset. Inspired by UOD mechanism and inlier priority phenomenon, where intuitively models fit inliers more quickly than outliers, we propose GradStop, a sampling-based label-free algorithm to estimate model's real-time performance during training. First, a sampling method generates two sets: one likely containing more outliers and the other more inliers, then a metric based on gradient cohesion is applied to probe into current training dynamics, which reflects model's performance on OD task. Experimental results on 4 deep UOD algorithms and 47 real-world datasets and theoretical proofs demonstrate the effectiveness of our proposed early stopping algorithm in enhancing the performance of deep UOD models. Auto Encoder (AE) enhanced by GradStop achieves better performance than itself, other SOTA UOD methods, and even ensemble AEs. Our method provides a robust and effective solution to the problem of performance degradation during training, enabling deep UOD models to achieve better potential in anomaly detection tasks.

cs.LG↗

arXiv2025

Neural Predictor for Flight Control with Payload

Aerial robotics for transporting suspended payloads as the form of freely-floating manipulator are growing great interest in recent years. However, the force/torque caused by payload and residual dynamics will introduce unmodeled perturbations to the aerial robotics, which negatively affects the closed-loop performance. Different from estimation-like methods, this paper proposes Neural Predictor, a learning-based approach to model force/torque induced by payload and residual dynamics as a dynamical system. It yields a hybrid model that combines the first-principles dynamics with the learned dynamics. The hybrid model is then integrated into a MPC framework to improve closed-loop performance. Effectiveness of proposed framework is verified extensively in both numerical simulations and real-world flight experiments. The results indicate that our approach can capture force/torque caused by suspended payload and residual dynamics accurately, respond quickly to the changes of them and improve the closed-loop performance significantly. In particular, Neural Predictor outperforms a state-of-the-art learning-based estimator and has reduced the force and torque estimation errors by up to 66.15% and 33.33% while requiring less samples. The code of proposed Neural Predictor can be found at https://github.com/NPU-RCIR/Neural-Predictor.git.

cs.RO↗

arXiv2025

GECAM Discovery of Peculiar Oscillating Particle Precipitation Events

Charged particle precipitation typically manifests as a gradual increase and decrease of flux observed by space detectors. Cases with rapidly flux variation are very rare. Periodic events are even more extraordinary. These oscillating particle precipitation (OPP) events are usually attributed to the bounce motion of electrons, which are induced by lightning. Owing to the observation limitations, there has been debate regarding whether these oscillations originate from temporal flux evolution or spatial structure evolution. Here we report three peculiar charged particle precipitation events detected by GECAM during a geomagnetic storm on March 21, 2024, with two exhibiting significant periodicity. These events were observed around the same region during three consecutive orbits. Through comprehensive temporal and spectral analyses, we revealed that one of the OPP events exhibited a transition in spectral lag of mini-pulses, shifting from "softer-earlier" to "softer-later" while showing no significant time evolution in overall frequency characteristics. And there is no association found between these two OPP events and lightning activity. Several possible scenarios are discussed to explain these charged particles with a life time of more than 3.5 hours, but the nature of these three events remains an enigma. We suggest that these GECAM-detected OPP events may represent a new type of particle precipitation event or a peculiar Lightning-induced Electron Precipitations (LEPs).

astro-ph.HE↗

arXiv2025

CellVerse: Do Large Language Models Really Understand Cell Biology?

Recent studies have demonstrated the feasibility of modeling single-cell data as natural languages and the potential of leveraging powerful large language models (LLMs) for understanding cell biology. However, a comprehensive evaluation of LLMs' performance on language-driven single-cell analysis tasks still remains unexplored. Motivated by this challenge, we introduce CellVerse, a unified language-centric question-answering benchmark that integrates four types of single-cell multi-omics data and encompasses three hierarchical levels of single-cell analysis tasks: cell type annotation (cell-level), drug response prediction (drug-level), and perturbation analysis (gene-level). Going beyond this, we systematically evaluate the performance across 14 open-source and closed-source LLMs ranging from 160M to 671B on CellVerse. Remarkably, the experimental results reveal: (1) Existing specialist models (C2S-Pythia) fail to make reasonable decisions across all sub-tasks within CellVerse, while generalist models such as Qwen, Llama, GPT, and DeepSeek family models exhibit preliminary understanding capabilities within the realm of cell biology. (2) The performance of current LLMs falls short of expectations and has substantial room for improvement. Notably, in the widely studied drug response prediction task, none of the evaluated LLMs demonstrate significant performance improvement over random guessing. CellVerse offers the first large-scale empirical demonstration that significant challenges still remain in applying LLMs to cell biology. By introducing CellVerse, we lay the foundation for advancing cell biology through natural languages and hope this paradigm could facilitate next-generation single-cell analysis.

q-bio.QM↗

arXiv2025

Enhanced Robust Tracking Control: An Online Learning Approach

This work focuses the tracking control problem for nonlinear systems subjected to unknown external disturbances. Inspired by contraction theory, a neural network-dirven CCM synthesis is adopted to obtain a feedback controller that could track any feasible trajectory. Based on the observation that the system states under continuous control input inherently contain embedded information about unknown external disturbances, we propose an online learning scheme that captures the disturbances dyanmics from online historical data and embeds the compensation within the CCM controller. The proposed scheme operates as a plug-and-play module that intrinsically enhances the tracking performance of CCM synthesis. The numerical simulations on tethered space robot and PVTOL demonstrate the effectiveness of proposed scheme. The source code of the proposed online learning scheme can be found at https://github.com/NPU-RCIR/Online_CCM.git.

eess.SY↗

arXiv2025

HOT3D: Hand and Object Tracking in 3D from Egocentric Multi-View Videos

We introduce HOT3D, a publicly available dataset for egocentric hand and object tracking in 3D. The dataset offers over 833 minutes (3.7M+ images) of recordings that feature 19 subjects interacting with 33 diverse rigid objects. In addition to simple pick-up, observe, and put-down actions, the subjects perform actions typical for a kitchen, office, and living room environment. The recordings include multiple synchronized data streams containing egocentric multi-view RGB/monochrome images, eye gaze signal, scene point clouds, and 3D poses of cameras, hands, and objects. The dataset is recorded with two headsets from Meta: Project Aria, which is a research prototype of AI glasses, and Quest 3, a virtual-reality headset that has shipped millions of units. Ground-truth poses were obtained by a motion-capture system using small optical markers attached to hands and objects. Hand annotations are provided in the UmeTrack and MANO formats, and objects are represented by 3D meshes with PBR materials obtained by an in-house scanner. In our experiments, we demonstrate the effectiveness of multi-view egocentric data for three popular tasks: 3D hand tracking, model-based 6DoF object pose estimation, and 3D lifting of unknown in-hand objects. The evaluated multi-view methods, whose benchmarking is uniquely enabled by HOT3D, significantly outperform their single-view counterparts.

cs.CV↗

arXiv2025

Decentralization of Ethereum's Builder Market

Blockchains protect an ecosystem worth more than $500bn with strong security properties derived from the principle of decentralization. Is today's blockchain decentralized? In this paper, we empirically studied one of the least decentralized parts of Ethereum, its builder market. The builder market was introduced to fairly distribute Maximal Extractable Value (MEV) among validators and avoid validator centralization. As of the time of writing, two builders produced more than 85% of blocks in Ethereum, creating a concerning centralization factor. However, a common belief is that such centralization "is okay," arguing that builder centralization will not lead to validator centralization. In this empirical study, we quantify the significant proposer losses within the centralized builder market and challenge the belief that this is acceptable. The significant proposer losses, if left uncontrolled, could undermine the goal of PBS. Moreover, MEV mitigation solutions slated for adoption are affected too because they rely on the builder market as an "MEV oracle," which is made inaccurate by centralization. Our investigation reveals the incentive issue within the current MEV supply chain and its implications for builder centralization and proposer losses. Finally, we analyze why the proposed mitigation cannot work and highlight two properties essential for effective solutions.

cs.CR↗

arXiv2025

UltraFusion: Ultra High Dynamic Imaging using Exposure Fusion

Capturing high dynamic range (HDR) scenes is one of the most important issues in camera design. Majority of cameras use exposure fusion, which fuses images captured by different exposure levels, to increase dynamic range. However, this approach can only handle images with limited exposure difference, normally 3-4 stops. When applying to very high dynamic range scenes where a large exposure difference is required, this approach often fails due to incorrect alignment or inconsistent lighting between inputs, or tone mapping artifacts. In this work, we propose \model, the first exposure fusion technique that can merge inputs with 9 stops differences. The key idea is that we model exposure fusion as a guided inpainting problem, where the under-exposed image is used as a guidance to fill the missing information of over-exposed highlights in the over-exposed region. Using an under-exposed image as a soft guidance, instead of a hard constraint, our model is robust to potential alignment issue or lighting variations. Moreover, by utilizing the image prior of the generative model, our model also generates natural tone mapping, even for very high-dynamic range scenes. Our approach outperforms HDR-Transformer on latest HDR benchmarks. Moreover, to test its performance in ultra high dynamic range scenes, we capture a new real-world exposure fusion benchmark, UltraFusion dataset, with exposure differences up to 9 stops, and experiments show that UltraFusion can generate beautiful and high-quality fusion results under various scenarios. Code and data will be available at https://openimaginglab.github.io/UltraFusion.

cs.CV↗

arXiv2025

Cross-domain Fiber Cluster Shape Analysis for Language Performance Cognitive Score Prediction

Shape plays an important role in computer graphics, offering informative features to convey an object's morphology and functionality. Shape analysis in brain imaging can help interpret structural and functionality correlations of the human brain. In this work, we investigate the shape of the brain's 3D white matter connections and its potential predictive relationship to human cognitive function. We reconstruct brain connections as sequences of 3D points using diffusion magnetic resonance imaging (dMRI) tractography. To describe each connection, we extract 12 shape descriptors in addition to traditional dMRI connectivity and tissue microstructure features. We introduce a novel framework, Shape--fused Fiber Cluster Transformer (SFFormer), that leverages a multi-head cross-attention feature fusion module to predict subject-specific language performance based on dMRI tractography. We assess the performance of the method on a large dataset including 1065 healthy young adults. The results demonstrate that both the transformer-based SFFormer model and its inter/intra feature fusion with shape, microstructure, and connectivity are informative, and together, they improve the prediction of subject-specific language performance scores. Overall, our results indicate that the shape of the brain's connections is predictive of human language function.

cs.CV↗

arXiv2025

TxGemma: Efficient and Agentic LLMs for Therapeutics

Therapeutic development is a costly and high-risk endeavor that is often plagued by high failure rates. To address this, we introduce TxGemma, a suite of efficient, generalist large language models (LLMs) capable of therapeutic property prediction as well as interactive reasoning and explainability. Unlike task-specific models, TxGemma synthesizes information from diverse sources, enabling broad application across the therapeutic development pipeline. The suite includes 2B, 9B, and 27B parameter models, fine-tuned from Gemma-2 on a comprehensive dataset of small molecules, proteins, nucleic acids, diseases, and cell lines. Across 66 therapeutic development tasks, TxGemma achieved superior or comparable performance to the state-of-the-art generalist model on 64 (superior on 45), and against state-of-the-art specialist models on 50 (superior on 26). Fine-tuning TxGemma models on therapeutic downstream tasks, such as clinical trial adverse event prediction, requires less training data than fine-tuning base LLMs, making TxGemma suitable for data-limited applications. Beyond these predictive capabilities, TxGemma features conversational models that bridge the gap between general LLMs and specialized property predictors. These allow scientists to interact in natural language, provide mechanistic reasoning for predictions based on molecular structure, and engage in scientific discussions. Building on this, we further introduce Agentic-Tx, a generalist therapeutic agentic system powered by Gemini 2.5 that reasons, acts, manages diverse workflows, and acquires external domain knowledge. Agentic-Tx surpasses prior leading models on the Humanity's Last Exam benchmark (Chemistry & Biology) with 52.3% relative improvement over o3-mini (high) and 26.7% over o3-mini (high) on GPQA (Chemistry) and excels with improvements of 6.3% (ChemBench-Preference) and 2.4% (ChemBench-Mini) over o3-mini (high).

cs.AI↗

arXiv2025

Agglomerating Large Vision Encoders via Distillation for VFSS Segmentation

The deployment of foundation models for medical imaging has demonstrated considerable success. However, their training overheads associated with downstream tasks remain substantial due to the size of the image encoders employed, and the inference complexity is also significantly high. Although lightweight variants have been obtained for these foundation models, their performance is constrained by their limited model capacity and suboptimal training strategies. In order to achieve an improved tradeoff between complexity and performance, we propose a new framework to improve the performance of low complexity models via knowledge distillation from multiple large medical foundation models (e.g., MedSAM, RAD-DINO, MedCLIP), each specializing in different vision tasks, with the goal to effectively bridge the performance gap for medical image segmentation tasks. The agglomerated model demonstrates superior generalization across 12 segmentation tasks, whereas specialized models require explicit training for each task. Our approach achieved an average performance gain of 2\% in Dice coefficient compared to simple distillation.

cs.CV↗

arXiv2025

VISTA: Unsupervised 2D Temporal Dependency Representations for Time Series Anomaly Detection

Time Series Anomaly Detection (TSAD) is essential for uncovering rare and potentially harmful events in unlabeled time series data. Existing methods are highly dependent on clean, high-quality inputs, making them susceptible to noise and real-world imperfections. Additionally, intricate temporal relationships in time series data are often inadequately captured in traditional 1D representations, leading to suboptimal modeling of dependencies. We introduce VISTA, a training-free, unsupervised TSAD algorithm designed to overcome these challenges. VISTA features three core modules: 1) Time Series Decomposition using Seasonal and Trend Decomposition via Loess (STL) to decompose noisy time series into trend, seasonal, and residual components; 2) Temporal Self-Attention, which transforms 1D time series into 2D temporal correlation matrices for richer dependency modeling and anomaly detection; and 3) Multivariate Temporal Aggregation, which uses a pretrained feature extractor to integrate cross-variable information into a unified, memory-efficient representation. VISTA's training-free approach enables rapid deployment and easy hyperparameter tuning, making it suitable for industrial applications. It achieves state-of-the-art performance on five multivariate TSAD benchmarks.

cs.LG↗