Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks138 from public data
- PMR: Prototypical Modal Rebalance for Multimodal Learning249
Prototypical Modality Rebalance (PMR) is proposed to perform stimulation on the particular slow-learning modality without interference from other modalities, and only relies on the representations of each modality and without restrictions from model structures and fusion methods, making it with great application potential for various scenarios.
- Self-Introspective Decoding: Alleviating Hallucinations for Large Vision-Language Models118
This work develops the Context and Text-aware Token Selection (CT2S) strategy, which preserves only unimportant vision tokens after early layers of LVLMs to adaptively amplify text-informed hallucination during the auto-regressive decoding.
- 117
- 102
- Towards Efficient Replay in Federated Incremental Learning59
This work proposes to employ a simple, generic frame-work for FIL named Re-Fed, which can coordinate each client to cache important samples for replay, alleviating the catastrophic forgetting problem in Federated Incremental Learning scenarios where edge clients may lack enough storage space to retain full data.
- 54
- Unleashing the Power of Continual Learning on Non-Centralized Devices: A Survey53
A large-scale benchmark is established to revisit this problem and analyze the performance of the state-of-the-art NCCL approaches and the important challenges and future research directions in NCCL are discussed.
- 51
- 50
- 49
- FedTune: A Deep Dive into Efficient Federated Fine-Tuning with Pre-trained Transformers47
It is demonstrated that the fine-tuned Transformers achieve extraordinary performance on FL, and that the lightweight fine- Tuning method facilitates a fast convergence rate and low communication costs.
- FedDSE: Distribution-aware Sub-model Extraction for Federated Learning over Resource-constrained Devices45
This paper identifies that when making predictions, different clients tend to activate different neurons of the entire model related to their respective distributions, and proposes a novel method called FedDSE, which can reduce the conflicts among clients by extracting sub-models based on the data distribution of each client.
- Aligning Language Models with Human Preferences via a Bayesian Approach42
A novel approach is proposed, which employs a Bayesian framework to account for the distribution of disagreements among human preferences as training a preference model, and names it as d-PM, which consistently exceeds previous SOTA models in both automatic and human evaluations.
- 41
- FedNLR: Federated Learning with Neuron-wise Learning Rates40
This paper proposes a novel and simple algorithm called FedNLR, which utilizes Neuron-wise Learning Rates during the FL local training process to enhance the learning of neurons bound to local classes on local data knowledge while reducing the decay of non-local classes knowledge stored in neurons.
- 39
- Edge Learning38
A comprehensive survey of the recent research efforts on Edge Learning, which provides an overview of the enabling technologies for EL, including model training, inference, security guarantee, privacy protection, and incentive mechanism.
- Personalized Federated Recommendation for Cold-Start Users via Adaptive Knowledge Fusion35
This work takes a dive into the federated recommendation for cold-start users, a novel scenario where the new participating users can directly obtain a promising recommendation without comprehensive training with all participating users by leveraging both transferred knowledge from the converged warm clients and the knowledge learned from the local data.
- FR: Folded Rationalization with a Unified Encoder33
Folded Rationalization (FR) is proposed that folds the two phases of the rationale model into one from the perspective of text semantic extraction to facilitate a better predictor by access to valuable information blocked by the generator in the traditional two-phase model and thus bring a better generator.
- 33
- Non-exemplar Online Class-Incremental Continual Learning via Dual-Prototype Self-Augment and Refinement29
A novel Dual-prototype Self-augment and Refinement method (DSR) for NO-CL problem, which consists of two strategies: Dual class prototypes: vanilla and high-dimensional prototypes are exploited to utilize the pre-trained information and obtain robust quasi-orthogonal representations rather than example buffers for both privacy preservation and memory reduction.
- D-Separation for Causal Self-Explanation29
This study proposes a novel criterion to uncover the causal rationale, termed the Minimum Conditional Dependence (MCD) criterion, and employs a simple and practical measure of dependence, specifically the KL-divergence, to validate the proposed MCD criterion.
- DCSpell: A Detector-Corrector Framework for Chinese Spelling Error Correction28
A cloze-style detector-corrector framework (DCSpell) that firstly detects whether a character is erroneous before correcting it, and employs sentence pairs as the input, which potentially incorporates the knowledge of phonological and visual similarity.
- OSP28
A new synchronization parallel mechanism named overlap synchronization parallel (OSP), in which the waiting time is removed by conducting computation and communication in an overlapped manner, and it is theoretically proved that this mechanism could achieve the same convergence rate compared to the sequential SGD for non-convex problems.
- Heterogeneity-aware Gradient Coding for Straggler Tolerance26
A novel heterogeneity-aware gradient coding scheme which can not only tolerate a predetermined number of stragglers but also fully utilize the computation capabilities of heterogenous workers is proposed.
- Stock Price Prediction Using Time Convolution Long Short-Term Memory Network24
A novel deep learning model named Time Convolution Long Short-Term Memory (TC-LSTM) networks is proposed to learn long-term dependencies of stock prices and outperforms other three baseline models in the mean square error.
- 23
- 22
- UNGER: Generative Recommendation with A Unified Code via Semantic and Collaborative Integration20
A novel method is proposed, named UNGER, which integrates semantic and collaborative knowledge into a UNified code for Generative Recommendation, and adaptively learn an integrated embedding through the joint optimization of cross-modality knowledge alignment and next-item prediction tasks.
- FedBAT: Communication-Efficient Federated Learning via Learnable Binarization20
This work proposes Federated Binarization-Aware Training (FedBAT), a novel framework that directly learns binary model updates during the local training process, thus inherently reducing the approximation errors.
- ProCC: Progressive Cross-Primitive Compatibility for Open-World Compositional Zero-Shot Learning19
This paper revisits the primitive prediction approach and proposes a novel method, termed Progressive Cross-primitive Compatibility (ProCC), to mimic the human learning process for OW-CZSL tasks, which outperforms other representative methods on both OW-CZSL and pCZSL settings by large margins.
- Decoupled Rationalization with Asymmetric Learning Rates: A Flexible Lipschitz Restraint19
The main idea of DR is to decouple the generator and predictor to allocate them with asymmetric learning rates, which can naturally and flexibly restrain the Lipschitz constant of the predictor, to address the problem of degeneration.
- Detached and Interactive Multimodal Learning18
DI-MML is introduced, a novel detached MML framework designed to learn complementary information across modalities under the premise of avoiding modality competition, and employs a certainty-aware logit weighting strategy to effectively leverage complementary information at the instance level during inference.
- Enhancing the Rationale-Input Alignment for Self-explaining Rationalization18
A novel approach called DAR (Discriminatively Aligned Rationalization), which utilizes an auxiliary module pretrained on the full input to discriminatively align the selected rationale and the original input, thereby overcoming the rationale shift problem.
- 18
- 18
- 17
- 17
- Deploying Foundation Model Powered Agent Services: A Survey16
A unified framework aimed at providing a comprehensive survey on deploying FM-based agent services across heterogeneous devices, with the emphasis on the integration of model and resource optimization to establish a robust infrastructure for these services is proposed.
- From Deterioration to Acceleration: A Calibration Approach to Rehabilitating Step Asynchronism in Federated Optimization16
A new algorithm FedaGrac is proposed, which calibrates the local direction to a predictive global orientation and holds an improved order of convergence rate than the state-of-the-art approaches and eliminates the negative effect of step asynchronism.
- 16
- Overcome Modal Bias in Multi-modal Federated Learning via Balanced Modality Selection15
This work introduces a modal enhancement loss during local training to alleviate local imbalance based on the aggregated global prototypes and proposes the modality selection aiming to select subsets of local modalities with great diversity and achieving global modal balance simultaneously.
- Knowledge-Aware Parameter Coaching for Personalized Federated Learning13
This paper proposes a knowledge-aware parameter coaching scheme where each client can swiftly and granularly refer to parameters of other clients to guide the local training, whereby accurate personalized client models can be efficiently produced without contradictory knowledge.
- Masked Random Noise for Communication-Efficient Federated Learning13
This paper proposes Federated Masked Random Noise (FedMRN), a novel framework that enables clients to learn a 1-bit mask for each model parameter and apply masked random noise to represent model updates to enhance communication efficiency.
- 13
- Transform-Equivariant Consistency Learning for Temporal Sentence Grounding13
A novel Equivariant Consistency Regulation Learning (ECRL) framework is introduced to learn more discriminative query-related frame-wise representations for each video, in a self-supervised manner, to address the temporal sentence grounding.
- Act-With-Think: Chunk Auto-Regressive Modeling for Generative Recommendation12
This work presents Chunk AutoRegressive Modeling (CAR), a new generation paradigm following the decision pattern that users usually think semantic aspects of items and then take actions on target items, which significantly outperforms existing methods based on traditional AR.
- 12
- Sign bit is enough12
Experimental results demonstrate that Marsit reduces up to 35% training time while preserving the same accuracy as training without compression, and the proposed framework retains the same theoretical convergence rate as non-compression mechanisms.
- The Panaceas for Improving Low-Rank Decomposition in Communication-Efficient Federated Learning10
This paper focuses on three key issues related to decomposition in FL: what to decompose, how to decompose, and how to aggregate, and introduces three novel techniques: Model Update Decomposition (MUD), Block-wise Kronecker Decomposition (BKD), and Aggregation-Aware Decomposition (AAD).
- VisionGuard: Secure and Robust Visual Perception of Autonomous Vehicles in Practice10
The key of VisionGuard is to leverage the spatiotemporal inconsistency property of PAEs to detect anomalies and it predicts the motion states from historical ones and compares them with the current driving states to identify any motion inconsistency caused by physical attacks.
- 10
- 10
- Better Knowledge Enhancement for Privacy‐Preserving Cross‐Project Defect Prediction9
A novel knowledge enhancement approach named FedDP is proposed with two simple but effective solutions: local heterogeneity awareness and global knowledge distillation that optimize the global model with the heterogeneity‐aware local model ensemble via knowledge distillation.
- 9
- 9
- 9
- BoRA: Towards More Expressive Low-Rank Adaptation with Block Diversity8
Block Diversified Low-Rank Adaptation is proposed, which improves the rank of LoRA weights with a small number of additional parameters and introduces a unique diagonal matrix for each block multiplication to enhance the diversity of different block products.
- Adversarial Cooperative Rationalization: The Risk of Spurious Correlations in Even Clean Datasets8
This study investigates the self-rationalization framework constructed with a cooperative game, where a generator initially extracts the most informative segment from raw input, and a subsequent predictor utilizes the selected subset for its input, and introduces an instruction to prevent the predictor from learning the correlations.
- OTAS: An Elastic Transformer Serving System via Token Adaptation8
This paper introduces OTAS, the first elastic serving system specially tailored for transformer models by exploring lightweight token management and implements and evaluates a prototype of OTAS with multiple datasets, which show that OTAS improves the system utility by at least 18.2%.
- 8
- A Systematic Survey on Federated Sequential Recommendation7
This survey pioneers Federated Sequential Recommendation (FedSR), where each user joins as a participant in federated training to achieve a recommendation service that balances data privacy and model performance.
- Is Model Editing Built on Sand? Revealing Its Illusory Success and Fragile Foundation7
Empirical evidence shows that editing is likely to be based on shortcuts rather than full semantics, calling for an urgent reconsideration of the very basis of model editing before further advancements can be meaningfully pursued.
- Mixed-Precision Embeddings for Large-Scale Recommendation Models7
This paper proposes a novel embedding compression method, termed Mixed-Precision Embeddings (MPE), which first group features by frequency and then search precision for each feature group, to reduce the size of the search space.
- Privacy-Friendly Cross-Domain Recommendation via Distilling User-irrelevant Information6
These prototypes facilitate bridging the domain shift between distinct source and target domains in a privacy-friendly manner and enable the identification of top-k users in the target domain to substitute for cold-start users prediction.
- Exploring Practical Gaps in Using Cross Entropy to Implement Maximum Mutual Information Criterion for Rationalization6
A simple yet effective method that introduces random vicinal1 perturbations to the selected rationale candidates and broadens the predictor’s assessment to a vicinity around the selected rationale candidate.
- Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligence6
A simple yet effective regularization algorithm for CFL named FedSSI is proposed, which tailors the synaptic intelligence for the CFL with heterogeneous data settings and can not only reduce computational overhead without rehearsal but also address the data heterogeneity issue.
- Balanced Multi-modal Federated Learning via Cross-Modal Infiltration6
This paper proposes a novel Cross-Modal Infiltration Federated Learning (FedCMI) framework that effectively alleviates modality imbalance and knowledge heterogeneity via knowledge transfer from the global dominant modality.
- FedBiF: Communication-Efficient Federated Learning via Bits Freezing5
This letter proposes Federated Bit Freezing (FedBiF), a novel FL framework that directly learns quantized model parameters during local training and attains accuracy comparable to FedAvg, even when using only 1 bit-per-parameter for uplink and 3 bpp for downlink communication.
- 5
- 5
- Disentangle Estimation of Causal Effects from Cross-Silo Data5
This work introduces an innovative disentangle architecture designed to facilitate the seamless cross-silo transmission of model parameters, enriched with causal mechanisms, through a combination of shared and private branches, thereby elevating the accuracy of the causal effect estimation.
- 5
- Self-Speculative Decoding for On-device MoE Acceleration4
This work proposes self-speculative MoE ( SS-MoE), an algorithm-system co-design framework that facilitates inference under limited GPU memory and significantly enhances decoding speed by 3.72x over state-of-the-art methods while maintaining nearly lossless accuracy.
- Multilevel Semantic-Aware Model for AI-Generated Video Quality Assessment4
MSA-VQA, a Multilevel Semantic-Aware Model for AI-Generated Video Quality Assessment, which leverages CLIP-based semantic supervision and cross-attention mechanisms is introduced, which achieves state-of-the-art results.
- 4
- HALO: Semantic-Aware Distributed LLM Inference in Lossy Edge Network3
HALO is a novel framework that can boost the distributed LLM inference in lossy edge network by strategically allocating less critical neuron groups to unstable devices, thus avoiding the excessive waiting time incurred by delayed packets.
- 3
- 3
- On the Convergence of Quantized Parallel Restarted SGD for Serverless Learning3
Quantized-PR-SGD is developed, a novel compression approach for serverless learning that integrates quantization and parallel restarted (PR) techniques to compress the exchanged information and to reduce synchronization frequency respectively, and the bound of accumulative errors according to synchronization mode and network topology is established.
- FedCD: Towards Consolidated Distillation for Heterogeneous Federated Learning2
The Consolidated Distillation for Heterogeneous Federated Learning - FedCD is proposed that balances knowledge representations from both feature-based and logit-based distillation to enhance performance and address the misalignment between knowledge conveyed by features and logits.
- Causality-inspired Federated Learning for Dynamic Spatio-Temporal Graphs2
A novel causality-inspired framework named SC-FSGL is proposed, which explicitly decouples transferable causal knowledge from client-specific noise through representation-level interventions and introduces a Conditional Separation Module that simulates soft interventions through client conditioned masks, enabling the disentanglement of invariant spatio-temporal causal factors from spurious signals.
- Ripple Shapley: Data Influence Attribution in One Federated Training Run2
Ripple Shapley is proposed, a novel attribution framework that enables accurate, real-time data valuation within a single federated training run and achieves up to 62× speedup over existing Shapley-based FL methods while maintaining high attribution fidelity, significantly improving efficiency, robustness, and fairness in federation environments.
- Enhancing RAG With Active Learning on Conversation Records: Reject Incapables and Answer Capables2
AL4RAG is introduced, a framework that uses active learning to select the most suitable conversation samples for annotation, filtering out redundant or noisy interactions and thereby optimizing model performance within a limited annotation budget.
- SNEAK: Synonymous Sentences-Aware Adversarial Attack on Natural Language Video Localization2
A new adversarial attack paradigm called synonymous sentences-aware adversarial attack on NLVL (SNEAK), which captures the cross-modality interplay between the vision and language sides is proposed, which captures the cross-modality interplay between the vision and language sides.
- 2
- FedRE: Robust and Effective Federated Learning with Privacy Preference2
The PSI within data is studied and the PSI within data is defined with regard to the privacy preferences of each client, and FedRE is developed, that can simultaneously achieve robustness and effectiveness benefits with LDP protection.
- 2
- Hypernetwork-driven centralized contrastive learning for federated graph classification2
This study introduces a novel hypernetwork-based method, termed CCL (Centralized Contrastive Learning), which is a server-centric innovation that effectively addresses the challenges posed by traditional client-centric approaches in heterogeneous datasets.
- 2
- 1
- 1
- KE-FedRS: Tackling Data Sparsity in Federated Recommendation via Knowledge Enhancement1
The core idea is to enhance the knowledge of users with few interactions and items with few ratings at both the local and global levels, and the proposed method consistently outperforms baseline approaches in terms of HR@10 and NDCG@10.
- 1
- 1
- PrismWF: A Multi-Granularity Patch-Based Transformer for Robust Website Fingerprinting Attack1
Pr PrismWF is proposed, a multi-granularity patch-based Transformer for multi-tab WF attack, which design a robust traffic feature representation for raw web traffic traces and extract multi-granularity features using convolutional kernels with different receptive fields to effectively integrate information across temporal scales.
- Efficient Knowledge Transfer in Federated Recommendation for Joint Venture Ecosystem1
This paper proposes an efficient and privacy-enhanced Federated Recommendation for the Joint Venture Ecosystem (FR-JVE) that each client transfers more common knowledge from other clients with a distilled user’s rating preference from the local dataset.
- Poster: Coda: Context-aware Acceleration for Distributed LLM Decoding in Edge1
Coda, a novel context-aware distributed acceleration framework tailored for packet loss scenarios, is proposed, which offers up to 5.62x speedup, significantly outperforming the state-of-the-art in various scenarios.
- 1
- 1
- 1
- 1
- 1
- LLM at Network Edge: A Layer-wise Efficient Federated Fine-tuning Approach1
Layer-wise Efficient Federated Fine-tuning (LEFF), a novel method designed to enhance the efficiency of FL fine-tuning while preserving model performance and minimizing client-side computational overhead, is introduced.
- 1
- 1
- 1
- FedGIG: Graph Inversion from Gradient in Federated Learning1
This paper first explores GIA's impact on Federated Graph Learning (FGL) and introduces Graph Inversion from Gradient in Federated Learning (FedGIG), a novel GIA method specifically designed for graph-structured data.
- DualMix: Unleashing the Potential of Data Augmentation for Online Class-Incremental Learning1
It is theoretically show that augmented samples with lower correlation to the original data are more effective in preventing forgetting and proposed the Enhanced Mixup (EnMix) method that mixes the augmented samples and their labels simultaneously, which is shown to enhance the sample diversity while maintaining strong consistency with corresponding labels.
- 1
- 1
- SimPattern: A Novel Byte-Pattern Matching based Chunk Similarity Detection for Post-Deduplication–
SimPattern is a similarity detection method motivated by the characteristics of delta compression that enables more fine-grained similarity detection and substantially improves the effectiveness of delta compression, reducing the final storage size by up to 42.68% compared with state-of-the-art methods.
- –
- –
- pFedDKS: Detached Knowledge Sharing for Personalized Federated Learning–
A substantial inconsistency is identified between the fused global feature representations and expected local feature representations, and thus it is necessary to preserve partially customized layers of the feature extractor for enhancing personalized representations.
- Data-Centric Sequential Recommendation with Relation-Augmented Generation–
This paper pioneers a relation-guided dataset augmentation and regeneration framework for sequential recommendation called RaSR, which can significantly improve model performance on original datasets while maintaining training efficiency without modifying the model architecture.
- –
- Take a Further Step Beyond Data Augmentation: Augmented Knowledge Purification for Federated Recommendation–
This paper proposes FR-AKP, which performs the augmented knowledge purification on the basis of data augmentation at both the intra-client and inter-client levels, and significantly outperforms state-of-the-art methods.
- FedKDD/FedMAS 2026: The 2026 International Joint Workshop on Federated Learning for Multi-agent Systems and Data Mining–
This workshop seeks to bring together researchers and practitioners from academia and industry to explore the convergence of federated learning and multi-agent systems, with a focus on scalability, trustworthiness, adaptive coordination, and the broader societal impact of federated multi-agent intelligence.
- –
- –
- Unbiased Rectification for Sequential Recommender Systems Under Fake Orders–
This paper identifies that fake orders are not absolutely harmful—in certain cases, partial fake orders can even have a data augmentation effect, and proposes Dual-view Identification and Targeted Rectification (DITaR), which primarily identifies harmful samples to achieve unbiased rectification of the system.
- –
- –
- –
- –
- –
- –
- ChatbotID: Identifying Chatbots with Granger Causality Test–
This work proposes an innovative approach named ChatbotID, which applies the Granger Causality Test (GCT) to extract a novel set of interactional features that capture the evolving, predictive relationships between conversational attributes, and significantly enhances the model’s ability to capture asymmetric influence in H-C dialogues.
- –
- –
- –
- –
- –
- –
- –
- –
- –
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.