Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Shaowei LinSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works137 from public data
- Scheduling Policies for Federated Learning in Wireless Networks671
An analytical model is developed to characterize the performance of federated learning in wireless networks and shows that running FL with PF outperforms RS and RR if the network is operating under a high signal-to-interference-plus-noise ratio (SINR) threshold, while RR is more preferable when the SINR threshold is low.
- An Unsupervised Sentence Embedding Method by Mutual Information Maximization197
Experimental results show that the proposed lightweight extension on top of BERT significantly outperforms other unsupervised sentence embedding baselines on common semantic textual similarity (STS) tasks and downstream supervised tasks, and achieves performance competitive with supervised methods on various tasks.
- Hulu-Med: A Transparent Generalist Model towards Holistic Medical Vision-Language Understanding113
Hulu-Med is introduced, a transparent, generalist medical Vision-Language Model (VLM) designed to unify language-only, 2D/3D vision-language, and video understanding within a single architecture, and outperforms proprietary systems such as GPT-4o on 16 benchmarks.
- Fast Model Debias with Machine Unlearning110
This work proposes a fast model debiasing framework (FMD) which offers an efficient approach to identify, evaluate and remove biases inherent in trained models and designs a machine unlearning-based strategy to efficiently and effectively remove the bias in a trained model with a small counterfactual dataset.
- SynCamMaster: Synchronizing Multi-Camera Video Generation from Diverse Viewpoints104
This work proposes a plug-and-play module that enhances a pre-trained text-to-video model for multi-camera video generation, ensuring consistent content across different viewpoints, and introduces a multi-view synchronization module to maintain appearance and geometry consistency across these viewpoints.
- Med-MoE: Mixture of Domain-Specific Experts for Lightweight Medical Vision-Language Models103
Comprehensive experiments on both open- and close-end medical question answering and image classification tasks across datasets such as VQA-RAD, SLAKE and Path-VQA demonstrate that the proposed Med-MoE model can achieve performance superior to or on par with state-of-the-art baselines, while only requiring approximately 30\%-50\% of activated model parameters.
- Deep learning-enabled 3D multimodal fusion of cone-beam CT and intraoral mesh scans for clinically applicable tooth-bone reconstruction85
Deep Dental Multimodal Fusion is presented, an automatic multimodal framework that reconstructs 3D tooth-bone structures using CBCT and IOS using a novel pixel representation learning architectures, prior knowledge-guided losses, and geometry-based 3D fusion techniques.
- Track without Appearance: Learn Box and Tracklet Embedding with Local and Global Motion Patterns for Vehicle Tracking75
A novel approach is proposed that tackles the association issue for long-term tracking with the exclusive fully-exploited motion information with the proposed reconstruct-to-embed strategy based on deep graph convolutional neural networks (GCN).
- TEaR: Improving LLM-based Machine Translation with Systematic Self-Refinement63
A systematic LLM-based self-refinement translation framework, named TEaR, which exhibits superior systematicity and interpretability, and conducts cross-model correction experiments to investigate the potential relationship between the translation and evaluation capabilities of general-purpose LLMs.
- 62
- UniEdit: A Unified Tuning-Free Framework for Video Motion and Appearance Editing61
UniEdit is presented, a tuning-free framework that supports both video motion and appearance editing by harnessing the power of a pre-trained text-to-video generator within an inversion-then-generation framework and introduces auxiliary motion-reference and reconstruction branches to produce text-guided motion and source features respectively.
- Comprehensive assessment of mRNA isoform detection methods for long-read sequencing data60
IsoQuant is demonstrated as a highly effective tool for isoform detection with LRS, with Bambu and StringTie2 also exhibiting strong performance and these results offer valuable guidance for future research on alternative splicing analysis and the ongoing improvement of tools for isoform detection using LRS data.
- 58
- 55
- A ChatGPT Aided Explainable Framework for Zero-Shot Medical Image Diagnosis52
This paper proposes a novel CLIP-based zero-shot medical image classification framework supplemented with ChatGPT for explainable diagnosis, mimicking the diagnostic process performed by human experts.
- 51
- FairMT-Bench: Benchmarking Fairness for Multi-turn Dialogue in Conversational LLMs50
A comprehensive fairness benchmark for LLMs in multi-turn dialogue scenarios, based on a task taxonomy targeting LLM fairness capabilities across three stages: context understanding, user interaction, and instruction trade-offs, is proposed.
- 50
- DIVOTrack: A Novel Dataset and Baseline Method for Cross-View Multi-Object Tracking in DIVerse Open Scenes49
The DIVOTrack is introduced: a novel baseline cross-view tracking method with a unified joint detection and cross-view tracking framework named CrossMOT, which learns object detection, single-view association, and cross-view matching with an all-in-one embedding model.
- Bootstrapped Unsupervised Sentence Representation Learning46
A new framework with a two-branch Siamese Network which maximizes the similarity between two augmented views of each sentence is proposed which significantly outperforms other state-of-the-art unsupervised methods on semantic textual similarity (STS) and classification tasks.
- Fed-GraB: Federated Long-tailed Learning with Self-Adjusting Gradient Balancer41
Using a Self-adjusting Gradient Balancer module that re-weights clients' gradients in a closed-loop manner, based on the feedback of global long-tailed distribution evaluated by a Direct Prior Analyzer module, clients can effectively alleviate the distribution drift caused by data heterogeneity during the model training process and obtain a global model with better performance on the minority classes while maintaining the performance of the majority classes.
- 37
- PAD: Personalized Alignment of LLMs at Decoding-Time36
Personalized Alignment at Decoding-time (PAD) is presented, a novel framework designed to align LLM outputs with diverse personalized preferences during the inference phase, eliminating the need for additional training.
- 36
- 34
- 33
- R-LLaVA: Improving Med-VQA Understanding through Visual Region of Interest33
R-LLaVA is introduced, designed to enhance biomedical VQA understanding by integrating simple medical annotations as prior knowledge directly into the image space through CLIP, and demonstrates R-LLaVA’s superiority over existing state-of-the-art (SoTA) methods.
- On the Effectiveness of Out-of-Distribution Data in Self-Supervised Long-Tail Learning31
This paper empirically identifies the counter-intuitive usefulness of OOD samples in SSL long-tailed learning and principally design a novel SSL method, Contrastive with Out-of-distribution (OOD) data for Long-Tail learning (COLT), which can effectively exploit OOD data to dynamically re-balance the feature space.
- TSegFormer: 3D Tooth Segmentation in Intraoral Scans with Geometry Guided Transformer30
This paper proposes TSegFormer which captures both local and global dependencies among different teeth and the gingiva in the IOS point clouds with a multi-task 3D transformer architecture, and designs a geometry-guided loss based on a novel point curvature to refine boundaries in an end-to-end manner.
- 29
- CP-Router: An Uncertainty-Aware Router Between LLM and LRM28
This work proposes CP-Router, a training-free, model-agnostic routing framework that dynamically selects between an LLM and an LRM, demonstrated with multiple-choice question answering (MCQA) prompts and introduces Full and Binary Entropy (FBE), a novel entropy-based criterion that adaptively selects the appropriate CP threshold.
- huARdb: human Antigen Receptor database for interactive clonotype-transcriptome analysis at the single-cell level26
The human Antigen Receptor database (huARdb), a large-scale human single-cell immune profiling database that contains 444 794 high confidence T or B cells with full-length TCR/BCR sequence and transcriptomes from 215 datasets, is presented.
- FAIntbench: A Holistic and Precise Benchmark for Bias Evaluation in Text-to-Image Models24
This study introduced FAIntbench, a holistic and precise benchmark for biases in T2I models, and applied it to evaluate seven recent large-scale T2I models and conducted human evaluation, whose results demonstrated the effectiveness of FAIntbench in identifying various biases.
- FedLoGe: Joint Local and Generic Federated Learning under Long-tailed Data24
This paper introduces an approach termed Federated Local and Generic Model Training in Fed-LT (FedLoGe), which enhances both local and generic model performance through the integration of representation learning and classifier alignment within a neural collapse framework.
- Towards Distribution-Agnostic Generalized Category Discovery24
A Self-Balanced Co-Advice contrastive framework (BaCon), which consists of a contrastive-learning branch and a pseudo-labeling branch, working collaboratively to provide interactive supervision to resolve the DA-GCD task, and compared with state-of-the-art methods from two closely related fields: imbalanced semi-supervised learning and generalized category discovery.
- A Transformer-Based Knowledge Distillation Network for Cortical Cataract Grading23
A novel Transformer-based Knowledge Distillation Network, called TKD-Net, for cortical cataract grading is proposed, which outperforms state-of-the-art methods, as well as the effectiveness of its key components.
- Large Language Model Bias Mitigation from the Perspective of Knowledge Editing21
A new bias mitigation benchmark BiasKE is established leveraging existing and additional constructed datasets, which systematically assesses debiasing performance by complementary metrics on fairness, specificity, and generalization, and a novel debiasing method, Fairness Stamp, is proposed, which enables editable fairness through fine-grained calibration on individual biased knowledge.
- MedThink: Explaining Medical Visual Question Answering via Multimodal Decision-Making Rationale20
A semi-automated annotation process to streamline data preparation and build new benchmark MedVQA datasets R-RAD, R-SLAKE and R-Path, which provide intermediate medical decision-making rationales generated by multimodal large language models and human annotations for question-answering pairs in existing MedVQA datasets.
- M-MAD: Multidimensional Multi-Agent Debate for Advanced Machine Translation Evaluation19
Comprehensive experiments show that M-MAD not only outperforms all existing LLM-as-a-judge methods but also competes with state-of-the-art reference-based automatic metrics, even when powered by a suboptimal model like GPT-4o mini.
- Detecting and classifying hidden defects in additively manufactured parts using deep learning and X-ray computed tomography19
How deep machine learning (ML) models trained using computer-generated images of defects can automatically identify defects in CT images of real parts that were never previously seen by the model is shown.
- Lightweight, Dynamic Graph Convolutional Networks for AMR-to-Text Generation19
This paper proposes Lightweight Dynamic Graph Convolutional Networks (LDGCNs) that capture richer non-local interactions by synthesizing higher order information from the input graphs and develops two novel parameter saving strategies based on the group graph convolutions and weight tied convolutions to reduce memory usage and model complexity.
- 19
- Joint Visual and Text Prompting for Improved Object-Centric Perception with Multimodal Large Language Models17
A novel approach, Joint Visual and Text Prompting (VTPrompt), that employs fine-grained visual information to enhance the capability of MLLMs in VQA, especially for object-oriented perception.
- Towards Federated Long-Tailed Learning17
This paper characterize three scenarios with different local or global long-tailed data distributions in the FL framework, and highlights the corresponding challenges.
- 17
- Identifying and Mitigating Social Bias Knowledge in Language Models16
A novel debiasing approach, Fairness Stamp (FAST), is proposed, which enables fine-grained calibration of individual social biases and surpasses state-of-the-art baselines with superior debiasing performance while not compromising the overall model capability for knowledge retention and downstream predictions.
- 15
- 15
- 15
- 14
- A Fine-Grained Attention Model for High Accuracy Operational Robot Guidance14
A low-cost edge computing-based IoT system based on an innovative fine-grained attention model (FGAM) that has shown superior performance over multiple benchmark models when validated using operational data and has been deployed on a welding robot in a real-world factory for mass production.
- 14
- Decoupling Dark Knowledge via Block-Wise Logit Distillation for Feature-Level Alignment13
A block-wise logit distillation framework to apply implicit logit-based feature alignment by gradually replacing teacher's blocks as intermediate stepping-stone models to bridge the gap between the student and the teacher is introduced.
- 13
- SurgBench: A Unified Large-Scale Benchmark for Surgical Video Analysis12
A unified surgical video benchmarking framework comprising a pretraining dataset, SurgBench-P, and an evaluation benchmark, SurgBench-E, that yields substantial performance improvements and superior cross-domain generalization to unseen procedures and modalities is introduced.
- 11
- Renovate Yourself: Calibrating Feature Representation of Misclassified Pixels for Semantic Segmentation11
The auxiliary representation calibration head (RCH) is proposed, which consists of the image decoupling, prototype clustering, error calibration modules and a metric loss function, to calibrate these error-prone feature representations for better intra-class consistency and segmentation performance.
- PaRT: Enhancing Proactive Social Chatbots with Personalized Real-Time Retrieval10
PaRT is presented, a novel framework enabling context-aware proactive dialogues for social chatbots through personalized real-time retrieval and generation of knowledge-grounded and engagement-optimized responses.
- MedEthicsQA: A Comprehensive Question Answering Benchmark for Medical Ethics Evaluation of LLMs10
A comprehensive benchmark comprising multiple-choice questions and open-ended questions for evaluation of medical ethics in LLMs is introduced and a hierarchical taxonomy integrating global medical ethical standards is systematically established.
- 10
- Beyond Modality Collapse: Representations Blending for Multimodal Dataset Distillation9
RepBlend is introduced, a novel MDD framework that weakens overdominant cross-modal supervision via representation blending via representation blending, thereby significantly enhancing intra-modal diversity and promoting balanced supervision and enhancing cross-modal alignment.
- 9
- 9
- Toothsegnet: Image Degradation Meets Tooth Segmentation in CBCT Images9
Experimental results suggest that ToothSegNet produces more precise segmentation and outperforms the state-of-the-art medical image segmentation methods.
- Fair-MoE: Fairness-Oriented Mixture of Experts in Vision-Language Models8
To build trust in medical VLMs, this work proposes Fair-MoE, a model specifically designed to ensure both fairness and effectiveness, and introduces a novel fairness-oriented loss function that optimizes the differences in the dispersion of various attributes' distributions.
- A pan-disease and population-level single-cell TCRαβ repertoire reference8
A large-scale reference of paired single-cell RNA/TCR sequencing comprising more than 2 million T cells from 70 studies, 1017 biological samples, 583 individuals, and 46 disease conditions is constructed, and TCR-DeepInsight, a computational framework to identify HLA-shared and disease-associated TCRαβ clusters that exhibit similar TCR sequence and GEX profiles is introduced.
- VersusDebias: Universal Zero-Shot Debiasing for Text-to-Image Models via SLM-Based Prompt Engineering and Generative Adversary8
VersusDebias is a novel and universal debiasing framework for biases in arbitrary T2I models, consisting of an array generation (AG) module and an image generation (IG) module, which has the capability to debias any models across gender, race, and age simultaneously simultaneously.
- TFormer: 3D Tooth Segmentation in Mesh Scans with Geometry Guided Transformer8
This paper proposes a novel method based on 3D transformer architectures that is evaluated with large-scale and high-resolution 3D IOS datasets, and captures both local and global dependencies among different teeth to distinguish various types of teeth with divergent anatomical structures and confusing boundaries.
- Federated Stochastic Gradient Descent Begets Self-Induced Momentum8
It is shown that running stochastic gradient descent in a federated learning setting can be viewed as adding a momentum-like term to the global aggregation process, which forges a link between staleness analysis and federated computing systems, which can be useful for systems designers.
- DynamicNER: A Dynamic, Multilingual, and Fine-Grained Dataset for LLM-based Named Entity Recognition7
DynamicNER is proposed, the first NER dataset designed for LLM-based methods with dynamic categorization, introducing various entity types and entity type lists for the same entity in different context, leveraging the generalization of LLM-based NER better.
- 7
- 7
- PX2Tooth: Reconstructing the 3D Point Cloud Teeth from a Single Panoramic X-Ray7
PX2Tooth, a novel approach to reconstruct 3D teeth using a single PX image with a two-stage framework, is proposed and a novel tooth generation network (TGNet) that learns to transform random point clouds into 3D teeth is designed.
- 7
- 7
- CAPO: Reinforcing Consistent Reasoning in Medical Decision-Making6
A novel large-scale RL framework for Med-VLMs, Consistency-Aware Preference Optimization (CAPO), which integrates rewards to ensure fidelity between perception and reasoning, consistency in reasoning-to-answer derivation, and rule-based accuracy for final responses is proposed.
- 6
- 6
- 6
- DentalBench: Benchmarking and Advancing LLMs Capability for Bilingual Dentistry Understanding5
DentalBench is introduced, the first comprehensive bilingual benchmark designed to evaluate and advance LLMs in the dental domain, and the importance of domain-specific benchmarks for developing trustworthy and effective LLMs tailored to healthcare applications is highlighted.
- 5
- MT$^{3}$: Scaling MLLM-based Text Image Machine Translation via Multi-Task Reinforcement Learning5
MT3, the first framework to apply Multi-Task RL to MLLMs for end-to-end TIMT, adopts a multi-task optimization paradigm targeting three key sub-skills: text recognition, context-aware reasoning, and translation, revealing how multi-task synergy, reinforcement learning initialization, curriculum design, and reward formulation contribute to advancing MLLM-driven TIMT.
- DCR: Divide-and-Conquer Reasoning for Multi-choice Question Answering with LLMs5
This work proposes to apply the Divide and Conquer to LLMs reasoning, which significantly boosts the models’ reasoning abilities across nine datasets involving arithmetic, commonsense, and logic tasks.
- Accurate estimation of 6-DoF tooth pose in 3D intraoral scans for dental applications using deep learning5
This paper proposes a novel method for automatic tooth orientation estimation by formulating it as a six-degree-of-freedom (6-DoF) tooth pose estimation task, and designs a deep neural network with a feature extractor backbone and a two-branch estimation head for tooth pose estimation.
- Cross-center Model Adaptive Tooth segmentation5
Experiments under three cross-center scenarios on two datasets show that CMAT can consistently surpass existing baselines, demonstrating its applicability for privacy-preserving model adaptive tooth segmentation in real-world digital dentistry.
- Privacy-Preserved Federated Learning for 3D Tooth Segmentation in Intra-Oral Mesh Scans5
This work proposes the FedTSeg framework, a federated 3D tooth segmentation framework with a deep graph convolutional neural network, and adopts a general privacy-preserving mechanism with homomorphic encryption to prevent information leakage during parameter exchange between the central server and local clients.
- 4
- Multi-class brain malignant tumor diagnosis in magnetic resonance imaging using convolutional neural networks4
Through multiple validations, FoTNet has proven to be an effective and robust tool for accurately classifying these brain tumors, providing strong support for preoperative diagnosis and assisting clinicians in making more informed treatment decisions.
- 4
- 4
- Clinically Applicable System For 3D Teeth Segmentation in Intraoral Scans using Deep Learning4
An accurate, efficient, and fully-automated deep learning model, trained on a dataset of 4,000 IOS data annotated by experienced human experts, demonstrating the potential for deep learning to improve the efficacy and efficiency of dental treatment and digital dentistry.
- Variational Probability Flow for Biologically Plausible Training of Deep Neural Networks4
It is shown that weight updates in VPF are local, depending only on the states and firing rates of the adjacent neurons, and, interestingly, if an asymmetric version of VPF exists, the weight updates directly explain experimental results in Spike-Timing-Dependent Plasticity (STDP).
- 4
- Pet-Bench: Benchmarking the Abilities of Large Language Models as E-Pets in Social Network Services3
PET-BENCH is introduced, a dedicated benchmark that evaluates Large Language Models across both self-interaction and human-interaction dimensions and serves as a foundational resource for benchmarking pet-related LLM abilities and advancing emotionally immersive human-pet interactions.
- LongWeave: A Long-Form Generation Benchmark Bridging Real-World Relevance and Verifiability3
LongWeave is introduced, which balances real-world and verifiable assessment with Constraint-Verifier Evaluation (CoV-Eval), and ensures that tasks are both realistic and objectively assessable, enabling rigorous assessment of model capabilities in meeting complex real-world constraints.
- 3
- 3
- 3
- 3
- Developing a universal and reliable temporomandibular joint reference system for quantifying morphological and positional changes3
This study provides a universal and reliable reference system for TMJ assessment that is applicable to both limited and large FOV CBCT and would improve comparability among diverse studies and enable comprehensive evaluations of TMJ positional and morphological changes during TMJ-related treatment follow-up such as splint therapy and disease progression.
- 3
- A multimodal vision-language model for comprehensive dental diagnosis and enhanced clinical practice2
DentVLM is a dental vision-language model developed to support dental diagnosis across seven oral imaging modalities and 36 tasks that surpasses junior readers, matches intermediate general practitioners and approaches senior specialists, and reduces diagnostic time by 15.0-37.0% in collaborative clinical workflows.
- 2
- 2
- 2
- EC-Guide: A Comprehensive E-Commerce Guide for Instruction Tuning and Quantization2
EC-Guide is a comprehensive e-commerce guide for instruction tuning and quantization of LLMs, enabling effective scalability across larger systems and heuristically integrated Chain-of-Thought during inference to enhance arithmetic performance.
- 2
- BIGbench: A Unified Benchmark for Evaluating Multi-dimensional Social Biases in Text-to-Image Models2
- 2
- 3D-U-SAM Network For Few-shot Tooth Segmentation in CBCT Images2
This work adopted a convolution approximation method in order to solve the problem of using 2D pre-trained weights on 3D datasets, and designed skip connections to fuse features at all levels with reference to U-Net to retain more details.
- Biologically Plausible Sequence Learning with Spiking Neural Networks2
It is shown that MPN can be trained to robustly memorize multiple spatiotemporal patterns of binary vectors, generalizing the ability of the symmetric Hopfield network to memorize static spatial patterns.
- GCAgent: Enhancing Group Chat Communication through Dialogue Agents System1
This work presents a practical blueprint for extending LLM-based dialogue agent from one-party chats to multi-party group scenarios, and introduces GCAgent, an LLM-driven system for enhancing group chats communication with both entertainment- and utility-oriented dialogue agents.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Heterogeneous sensor data fusion by deep learning1
This chapter provides an overview of heterogeneous sensor data fusion, including the background, basic deep learning techniques, and how these techniques can be used for sensorData fusion tasks.
- Bridging Model Heterogeneity in Federated Learning with Group Prototypical Alignment–
FedGPA is proposed, a hierarchical framework in which clients with comparable resources and identical architectures form a group and a server mediates knowledge transfer across groups of diverse models, and at its core, the lightweight, model-agnostic Aligned Co-decision (Alco) Unit aligns class-level prototypical information across groups to bridge heterogeneous architectures.
- –
- –
- HICT: High-Precision 3D CBCT Reconstruction From A Single X-ray–
HiCT is proposed, a two-stage framework that first generates geometrically consistent multi-view projections from a single panoramic image using a video diffusion model, and then reconstructs high-fidelity CBCT from the projections using a ray-based dynamic attention network and an X-ray sampling strategy.
- –
- FUSEP: A Multi-Center Benchmark for Diverse Tasks in Fetal Ultrasound Screening Early Pregnancy–
This work presents a benchmark dataset for Fetal Ultrasound Screening in Early Pregnancy to facilitate intelligent ultrasound examination and assisted diagnosis called FUSEP and believes that FUSEP and benchmark can contribute to the medical community in the development of multiple tasks such as standard plane recognition, quality control on ultrasound images, automated assisted diagnostics in early fetal pregnancy, medical multi-object detection, domain adaptation for object detection, etc.
- Modest-Align: Data-Efficient Alignment for Vision-Language Models–
Modest-Align is proposed, a lightweight alignment framework designed for robustness and efficiency that outperforms state-of-the-art methods in retrieval tasks, achieving competitive results with over 100x less training data and 600x less GPU time than CLIP.
- –
- –
- –
- –
- –
- –
- –
- Comprehensive Assessment of Isoform Detection Methods for Third-Generation Sequencing Data–
A benchmarking analysis of eleven methods implemented in eight computational tools capable of identifying isoform structures from TGS RNA sequencing data shows the guided mode of StringTie2 and Bambu achieved the best performance in sensitivity and precision.
- Tooth Defect Segmentation in 3D Mesh Scans Using Deep Learning–
A novel triple-stream graph convolutional network named TripleNet is proposed to learn multi-scale geometric features from mesh scans for end-to-end tooth defect segmentation in 3D mesh scans and achieves state-of-the-art performance on 3D tooth defect segmentation.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.