arXiv subjects

Fu Li

Publications and source records attributed to Fu Li.

At least 37 records · Page 2Linked to original sources

arXiv2025

Shift Current Anomalous Photovoltaics in a Double Perovskite Ferroelectric

Ferroelectric anomalous photovoltaic (APV) effect, as a fascinating physical conceptual phenomenon, holds significant potentials for new optoelectronic device applications. However, due to the knowledge lacking on the origin and underlying mechanism of ferroelectric APV effect, substantial challenges still remain in exploring new APV-active candidate materials. The emerging shift current model, involving the transfer of photogenerated charges through the displacement of wave functions, has attracted considerable attention for its unique insights into the bulk photovoltaic effect. Here, we present strong APV properties in a high-temperature double perovskite ferroelectric (cyclohexylmethylammonium)$_2$CsAgBiBr$_7$, showing an extremely large above-bandgap photovoltage up to about 40 V. This figure-of-merit is far beyond its bandgap of about 2.3 eV and comparable to the state-of-art molecular ferroelectrics. Strikingly, the shift current model reveals an intrinsic correlation with Cs$^{+}$ cation displacement and provides, for the first time, an explicit explanation for the structural origin of ferroelectric APV activities. Besides, its steady-state APV photocurrent exhibits the unique light-polarization dependence, which endows remarkable polarization-sensitivity with the highest polarization ratios of about 41 among the known 2D single-phase materials. As the unprecedented exploration of ferroelectric APV characteristics illuminated by the shift current mechanism, this finding paves a pathway to assemble new optoelectronic smart devices.

cond-mat.mtrl-sci↗

arXiv2025

Request-Only Optimization for Recommendation Systems

Deep Learning Recommendation Models (DLRMs) represent one of the largest machine learning applications on the planet. Industry-scale DLRMs are trained with petabytes of recommendation data to serve billions of users every day. To utilize the rich user signals in the long user history, DLRMs have been scaled up to unprecedented complexity, up to trillions of floating-point operations (TFLOPs) per example. This scale, coupled with the huge amount of training data, necessitates new storage and training algorithms to efficiently improve the quality of these complex recommendation systems. In this paper, we present a Request-Only Optimizations (ROO) training and modeling paradigm. ROO simultaneously improves the storage and training efficiency as well as the model quality of recommendation systems. We holistically approach this challenge through co-designing data (i.e., request-only data), infrastructure (i.e., request-only based data processing pipeline), and model architecture (i.e., request-only neural architectures). Our ROO training and modeling paradigm treats a user request as a unit of the training data. Compared with the established practice of treating a user impression as a unit, our new design achieves native feature deduplication in data logging, consequently saving data storage. Second, by de-duplicating computations and communications across multiple impressions in a request, this new paradigm enables highly scaled-up neural network architectures to better capture user interest signals, such as Generative Recommenders (GRs) and other request-only friendly architectures.

cs.IR↗

arXiv2025

DeltaEdit: Exploring Text-free Training for Text-Driven Image Manipulation

Text-driven image manipulation remains challenging in training or inference flexibility. Conditional generative models depend heavily on expensive annotated training data. Meanwhile, recent frameworks, which leverage pre-trained vision-language models, are limited by either per text-prompt optimization or inference-time hyper-parameters tuning. In this work, we propose a novel framework named \textit{DeltaEdit} to address these problems. Our key idea is to investigate and identify a space, namely delta image and text space that has well-aligned distribution between CLIP visual feature differences of two images and CLIP textual embedding differences of source and target texts. Based on the CLIP delta space, the DeltaEdit network is designed to map the CLIP visual features differences to the editing directions of StyleGAN at training phase. Then, in inference phase, DeltaEdit predicts the StyleGAN's editing directions from the differences of the CLIP textual features. In this way, DeltaEdit is trained in a text-free manner. Once trained, it can well generalize to various text prompts for zero-shot inference without bells and whistles.

cs.CV↗

arXiv2025

Diffusion-driven SpatioTemporal Graph KANsformer for Medical Examination Recommendation

Recommendation systems in AI-based medical diagnostics and treatment constitute a critical component of AI in healthcare. Although some studies have explored this area and made notable progress, healthcare recommendation systems remain in their nascent stage. And these researches mainly target the treatment process such as drug or disease recommendations. In addition to the treatment process, the diagnostic process, particularly determining which medical examinations are necessary to evaluate the condition, also urgently requires intelligent decision support. To bridge this gap, we first formalize the task of medical examination recommendations. Compared to traditional recommendations, the medical examination recommendation involves more complex interactions. This complexity arises from two folds: 1) The historical medical records for examination recommendations are heterogeneous and redundant, which makes the recommendation results susceptible to noise. 2) The correlation between the medical history of patients is often irregular, making it challenging to model spatiotemporal dependencies. Motivated by the above observation, we propose a novel Diffusion-driven SpatioTemporal Graph KANsformer for Medical Examination Recommendation (DST-GKAN) with a two-stage learning paradigm to solve the above challenges. In the first stage, we exploit a task-adaptive diffusion model to distill recommendation-oriented information by reducing the noises in heterogeneous medical data. In the second stage, a spatiotemporal graph KANsformer is proposed to simultaneously model the complex spatial and temporal relationships. Moreover, to facilitate the medical examination recommendation research, we introduce a comprehensive dataset. The experimental results demonstrate the state-of-the-art performance of the proposed method compared to various competitive baselines.

cs.IR↗

arXiv2025

CoCoDiff: Diversifying Skeleton Action Features via Coarse-Fine Text-Co-Guided Latent Diffusion

In action recognition tasks, feature diversity is essential for enhancing model generalization and performance. Existing methods typically promote feature diversity by expanding the training data in the sample space, which often leads to inefficiencies and semantic inconsistencies. To overcome these problems, we propose a novel Coarse-fine text co-guidance Diffusion model (CoCoDiff). CoCoDiff generates diverse yet semantically consistent features in the latent space by leveraging diffusion and multi-granularity textual guidance. Specifically, our approach feeds spatio-temporal features extracted from skeleton sequences into a latent diffusion model to generate diverse action representations. Meanwhile, we introduce a coarse-fine text co-guided strategy that leverages textual information from large language models (LLMs) to ensure semantic consistency between the generated features and the original inputs. It is noted that CoCoDiff operates as a plug-and-play auxiliary module during training, incurring no additional inference cost. Extensive experiments demonstrate that CoCoDiff achieves SOTA performance on skeleton-based action recognition benchmarks, including NTU RGB+D, NTU RGB+D 120 and Kinetics-Skeleton.

cs.CV↗

arXiv2025

Adaptive Progressive Attention Graph Neural Network for EEG Emotion Recognition

In recent years, numerous neuroscientific studies demonstrate that specific areas of the brain are connected to human emotional responses, with these regions exhibiting variability across individuals and emotional states. To fully leverage these neural patterns, we propose an Adaptive Progressive Attention Graph Neural Network (APAGNN), which dynamically captures the spatial relationships among brain regions during emotional processing. The APAGNN employs three specialized experts that progressively analyze brain topology. The first expert captures global brain patterns, the second focuses on region-specific features, and the third examines emotion-related channels. This hierarchical approach enables increasingly refined analysis of neural activity. Additionally, a weight generator integrates the outputs of all three experts, balancing their contributions to produce the final predictive label. Extensive experiments conducted on SEED, SEED-IV and MPED datasets indicate that our method enhances EEG emotion recognition performance, achieving superior results compared to baseline methods.

eess.SP↗

arXiv2025

Painting with Words: Elevating Detailed Image Captioning with Benchmark and Alignment Learning

Image captioning has long been a pivotal task in visual understanding, with recent advancements in vision-language models (VLMs) significantly enhancing the ability to generate detailed image captions. However, the evaluation of detailed image captioning remains underexplored due to outdated evaluation metrics and coarse annotations. In this paper, we introduce DeCapBench along with a novel metric, DCScore, specifically designed for detailed captioning tasks. DCScore evaluates hallucinations and fine-grained comprehensiveness by deconstructing responses into the smallest self-sufficient units, termed primitive information units, and assessing them individually. Our evaluation shows that DCScore aligns more closely with human judgment than other rule-based or model-based metrics. Concurrently, DeCapBench exhibits a high correlation with VLM arena results on descriptive tasks, surpassing existing benchmarks for vision-language models. Additionally, we present an automatic fine-grained feedback collection method, FeedQuill, for preference optimization based on our advanced metric, showing robust generalization capabilities across auto-generated preference data. Extensive experiments on multiple VLMs demonstrate that our method not only significantly reduces hallucinations but also enhances performance across various benchmarks, achieving superior detail captioning performance while surpassing GPT-4o.

cs.CV↗

arXiv2025

WVEmbs with its Masking: A Method For Radar Signal Sorting

Our study proposes a novel embedding method, Wide-Value-Embeddings (WVEmbs), for processing Pulse Descriptor Words (PDWs) as normalized inputs to neural networks. This method adapts to the distribution of interleaved radar signals, ranking original signal features from trivial to useful and stabilizing the learning process. To address the imbalance in radar signal interleaving, we introduce a value dimension masking method on WVEmbs, which automatically and efficiently generates challenging samples, and constructs interleaving scenarios, thereby compelling the model to learn robust features. Experimental results demonstrate that our method is an efficient end-to-end approach, achieving high-granularity, sample-level pulse sorting for high-density interleaved radar pulse sequences in complex and non-ideal environments.

eess.SP↗

arXiv2025

Learned Correction Methods for Ultrasound Computed Tomography Imaging Using Simplified Physics Models

Ultrasound computed tomography (USCT) is an emerging modality for breast imaging. Image reconstruction methods that incorporate accurate wave physics produce high resolution quantitative images of acoustic properties but are computationally expensive. The use of a simplified linear model in reconstruction reduces computational expense at the cost of reduced accuracy. This work aims to systematically compare different learning approaches for USCT reconstruction utilizing simplified linear models. This work considered various learning approaches to compensate for errors stemming from a linearized wave propagation model: correction in the data and image domains. The resulting image reconstruction methods are systematically assessed, alongside data-driven and model-based methods, in four virtual imaging studies utilizing anatomically realistic numerical phantoms. Image quality was assessed utilizing relative root mean square error (RRMSE), structural similarity index measure (SSIM), and a task-based assessment for tumor detection. Correction in the measurement domain resulted in images with minor visual artifacts and highly accurate task performance. Correction in the image domain demonstrated a heavy bias on training data, resulting in hallucinations, but greater robustness to measurement noise. Combining both forms of correction performed best in terms of RRMSE and SSIM, at the cost of task performance. This work systematically assessed learned reconstruction methods incorporating an approximated physical model for USCT imaging. Results demonstrated the importance of incorporating physics, compared to data-driven methods. Learning a correction in the data domain led to better task performance and robust out-of-distribution generalization compared to correction in the image domain.

eess.IV↗

arXiv2025

Goku: Flow Based Video Generative Foundation Models

This paper introduces Goku, a state-of-the-art family of joint image-and-video generation models leveraging rectified flow Transformers to achieve industry-leading performance. We detail the foundational elements enabling high-quality visual generation, including the data curation pipeline, model architecture design, flow formulation, and advanced infrastructure for efficient and robust large-scale training. The Goku models demonstrate superior performance in both qualitative and quantitative evaluations, setting new benchmarks across major tasks. Specifically, Goku achieves 0.76 on GenEval and 83.65 on DPG-Bench for text-to-image generation, and 84.85 on VBench for text-to-video tasks. We believe that this work provides valuable insights and practical advancements for the research community in developing joint image-and-video generation models.

cs.CV↗

arXiv2025

Spatio-Temporal Progressive Attention Model for EEG Classification in Rapid Serial Visual Presentation Task

As a type of multi-dimensional sequential data, the spatial and temporal dependencies of electroencephalogram (EEG) signals should be further investigated. Thus, in this paper, we propose a novel spatial-temporal progressive attention model (STPAM) to improve EEG classification in rapid serial visual presentation (RSVP) tasks. STPAM first adopts three distinct spatial experts to learn the spatial topological information of brain regions progressively, which is used to minimize the interference of irrelevant brain regions. Concretely, the former expert filters out EEG electrodes in the relative brain regions to be used as prior knowledge for the next expert, ensuring that the subsequent experts gradually focus their attention on information from significant EEG electrodes. This process strengthens the effect of the important brain regions. Then, based on the above-obtained feature sequence with spatial information, three temporal experts are adopted to capture the temporal dependence by progressively assigning attention to the crucial EEG slices. Except for the above EEG classification method, in this paper, we build a novel Infrared RSVP EEG Dataset (IRED) which is based on dim infrared images with small targets for the first time, and conduct extensive experiments on it. The results show that our STPAM can achieve better performance than all the compared methods.

cs.CV↗

arXiv2025

Integrating Language-Image Prior into EEG Decoding for Cross-Task Zero-Calibration RSVP-BCI

Rapid Serial Visual Presentation (RSVP)-based Brain-Computer Interface (BCI) is an effective technology used for information detection by detecting Event-Related Potentials (ERPs). The current RSVP decoding methods can perform well in decoding EEG signals within a single RSVP task, but their decoding performance significantly decreases when directly applied to different RSVP tasks without calibration data from the new tasks. This limits the rapid and efficient deployment of RSVP-BCI systems for detecting different categories of targets in various scenarios. To overcome this limitation, this study aims to enhance the cross-task zero-calibration RSVP decoding performance. First, we design three distinct RSVP tasks for target image retrieval and build an open-source dataset containing EEG signals and corresponding stimulus images. Then we propose an EEG with Language-Image Prior fusion Transformer (ELIPformer) for cross-task zero-calibration RSVP decoding. Specifically, we propose a prompt encoder based on the language-image pre-trained model to extract language-image features from task-specific prompts and stimulus images as prior knowledge for enhancing EEG decoding. A cross bidirectional attention mechanism is also adopted to facilitate the effective feature fusion and alignment between the EEG and language-image features. Extensive experiments demonstrate that the proposed model achieves superior performance in cross-task zero-calibration RSVP decoding, which promotes the RSVP-BCI system from research to practical application.

cs.HC↗

arXiv2024

Investigating the Use of Traveltime and Reflection Tomography for Deep Learning-Based Sound-Speed Estimation in Ultrasound Computed Tomography

Ultrasound computed tomography (USCT) quantifies acoustic tissue properties such as the speed-of-sound (SOS). Although full-waveform inversion (FWI) is an effective method for accurate SOS reconstruction, it can be computationally challenging for large-scale problems. Deep learning-based image-to-image learned reconstruction (IILR) methods can offer computationally efficient alternatives. This study investigates the impact of the chosen input modalities on IILR methods for high-resolution SOS reconstruction in USCT. The selected modalities are traveltime tomography (TT) and reflection tomography (RT), which produce a low-resolution SOS map and a reflectivity map, respectively. These modalities have been chosen for their lower computational cost relative to FWI and their capacity to provide complementary information: TT offers a direct SOS measure, while RT reveals tissue boundary information. Systematic analyses were facilitated by employing a virtual USCT imaging system with anatomically realistic numerical breast phantoms. Within this testbed, a supervised convolutional neural network (CNN) was trained to map dual-channel (TT and RT images) to a high-resolution SOS map. Single-input CNNs were trained separately using inputs from each modality alone (TT or RT) for comparison. The accuracy of the methods was systematically assessed using normalized root mean squared error (NRMSE), structural similarity index measure (SSIM), and peak signal-to-noise ratio (PSNR). For tumor detection performance, receiver operating characteristic analysis was employed. The dual-channel IILR method was also tested on clinical human breast data. Ensemble average of the NRMSE, SSIM, and PSNR evaluated on this clinical dataset were 0.2355, 0.8845, and 28.33 dB, respectively.

eess.IV↗

arXiv2024

Learning from the Web: Language Drives Weakly-Supervised Incremental Learning for Semantic Segmentation

Current weakly-supervised incremental learning for semantic segmentation (WILSS) approaches only consider replacing pixel-level annotations with image-level labels, while the training images are still from well-designed datasets. In this work, we argue that widely available web images can also be considered for the learning of new classes. To achieve this, firstly we introduce a strategy to select web images which are similar to previously seen examples in the latent space using a Fourier-based domain discriminator. Then, an effective caption-driven reharsal strategy is proposed to preserve previously learnt classes. To our knowledge, this is the first work to rely solely on web images for both the learning of new concepts and the preservation of the already learned ones in WILSS. Experimental results show that the proposed approach can reach state-of-the-art performances without using manually selected and annotated data in the incremental steps.

cs.CV↗

arXiv2024

Scaling of quantum Fisher information for quantum exceptional point sensors

In recent years, significant progress has been made in utilizing the divergence of spectrum response rate at the exceptional point (EP) for sensing in classical systems, while the use and characterization of quantum EPs for sensing have been largely unexplored. For a quantum EP sensor, an important issue is the relation between the order of the quantum EP and the scaling of quantum Fisher information (QFI), an essential quantity for characterizing quantum sensors. Here we investigate multi-mode quadratic bosonic systems, which exhibit higher-order EP dynamics, but possess Hermitian Hamiltonians without Langevin noise, thus can be utilized for quantum sensing. We derive an exact analytic formula for the QFI, from which we establish a scaling relation between the QFI and the order of the EP. We apply the formula to study a three-mode EP sensor and a multi-mode bosonic Kitaev chain and show that the EP physics can significantly enhance the sensing sensitivity. Our work establishes the connection between two important fields: non-Hermitian EP dynamics and quantum sensing, and may find important applications in quantum information and quantum non-Hermitian physics.

quant-ph↗

arXiv2024

Enhancing Cross-Dataset EEG Emotion Recognition: A Novel Approach with Emotional EEG Style Transfer Network

Recognizing the pivotal role of EEG emotion recognition in the development of affective Brain-Computer Interfaces (aBCIs), considerable research efforts have been dedicated to this field. While prior methods have demonstrated success in intra-subject EEG emotion recognition, a critical challenge persists in addressing the style mismatch between EEG signals from the source domain (training data) and the target domain (test data). To tackle the significant inter-domain differences in cross-dataset EEG emotion recognition, this paper introduces an innovative solution known as the Emotional EEG Style Transfer Network (E$^2$STN). The primary objective of this network is to effectively capture content information from the source domain and the style characteristics from the target domain, enabling the reconstruction of stylized EEG emotion representations. These representations prove highly beneficial in enhancing cross-dataset discriminative prediction. Concretely, E$^2$STN consists of three key modules\textemdash transfer module, transfer evaluation module, and discriminative prediction module\textemdash which address the domain style transfer, transfer quality evaluation, and discriminative prediction, respectively. Extensive experiments demonstrate that E$^2$STN achieves state-of-the-art performance in cross-dataset EEG emotion recognition tasks.

cs.HC↗

arXiv2024

A Study on Training and Developing Large Language Models for Behavior Tree Generation

This paper presents an innovative exploration of the application potential of large language models (LLM) in addressing the challenging task of automatically generating behavior trees (BTs) for complex tasks. The conventional manual BT generation method is inefficient and heavily reliant on domain expertise. On the other hand, existing automatic BT generation technologies encounter bottlenecks related to task complexity, model adaptability, and reliability. In order to overcome these challenges, we propose a novel methodology that leverages the robust representation and reasoning abilities of LLMs. The core contribution of this paper lies in the design of a BT generation framework based on LLM, which encompasses the entire process, from data synthesis and model training to application developing and data verification. Synthetic data is introduced to train the BT generation model (BTGen model), enhancing its understanding and adaptability to various complex tasks, thereby significantly improving its overall performance. In order to ensure the effectiveness and executability of the generated BTs, we emphasize the importance of data verification and introduce a multilevel verification strategy. Additionally, we explore a range of agent design and development schemes with LLM as the central element. We hope that the work in this paper may provide a reference for the researchers who are interested in BT generation based on LLMs.

cs.CL↗

arXiv2023

Retinex-guided Channel-grouping based Patch Swap for Arbitrary Style Transfer

The basic principle of the patch-matching based style transfer is to substitute the patches of the content image feature maps by the closest patches from the style image feature maps. Since the finite features harvested from one single aesthetic style image are inadequate to represent the rich textures of the content natural image, existing techniques treat the full-channel style feature patches as simple signal tensors and create new style feature patches via signal-level fusion, which ignore the implicit diversities existed in style features and thus fail for generating better stylised results. In this paper, we propose a Retinex theory guided, channel-grouping based patch swap technique to solve the above challenges. Channel-grouping strategy groups the style feature maps into surface and texture channels, which prevents the winner-takes-all problem. Retinex theory based decomposition controls a more stable channel code rate generation. In addition, we provide complementary fusion and multi-scale generation strategy to prevent unexpected black area and over-stylised results respectively. Experimental results demonstrate that the proposed method outperforms the existing techniques in providing more style-consistent textures while keeping the content fidelity.

cs.CV↗