Large Language Models (LLMs) have achieved remarkable success across reasoning and knowledge-intensive tasks, yet their static pretraining leaves them unable to handle rapidly evolving or domain-specific knowledge. Retrieval-Augmented Generation (RAG) addresses this by grounding LLM outputs in dynamically retrieved evidence, improving factual accuracy and reducing hallucinations. However, standard RAG pipelines struggle with temporally sensitive queries, especially when documents contain fuzzy or indirect time expressions (e.g., “a few years later”). This leads to Temporal Misalignment, where topically relevant but temporally incorrect results are retrieved. To overcome this, we propose DeFuzzRAG, a lightweight framework that enhances temporal robustness in RAG. DeFuzzRAG employs a small local language model to infer concrete time scopes from vague expressions and applies metadata-based filtering to realign retrieval with the query’s temporal intent. Experiments on a benchmark of fuzzified queries demonstrate that DeFuzzRAG substantially improves retrieval accuracy, raising Hit Rate by 15.7% while maintaining efficiency and model-agnostic integration. Our findings highlight the importance of temporal reasoning in RAG and establish DeFuzzRAG as a practical, plug-and-play solution for deploying temporally robust LLM systems in real-world settings.

Deploying Vision Transformers (ViTs) in real-world multi-task learning remains challenging due to their massive computational costs and the difficulty of pruning shared backbones without harming task performance. Single-task pruning often causes destructive interference by discarding weights critical to other tasks, while existing multi-task pruning strategies remain costly and unscalable for billion-parameter models. We propose Multi-LoRA Guided Importance Consensus (LoGIC), a unified framework for efficient and robust multi-task ViT pruning. LoGIC follows a two-phase procedure: (i) task-consistent pruning of LoRA modules, guided by a task-adaptive gating mechanism that balances shared and task-specific contributions while enforcing structured sparsity for deployment; and (ii) cross-task consensus pruning of the frozen ViT backbone, which retains both universally shared and task-specialized capabilities, enabling aggressive sparsity without sacrificing accuracy. Across five diverse vision benchmarks, LoGIC achieves up to 50% structured sparsity while maintaining competitive accuracy and surpassing all baselines.

Recently, Non-Fungible Tokens (NFTs) have attracted attention as valuable digital assets. However, NFT marketplaces face complex challenges in simultaneously recommending optimal pricing to sellers and desirable NFTs to buyers. Unlike conventional marketplaces that focus only on balancing demand and supply between sellers and buyers, these tasks are complicated by intricate value interdependencies arising from diverse buyer preferences, budgets, trait rarities, and the unprecedented breeding mechanisms. This paper formulates the NFT Project Pricing/Purchasing Recommendation (NP3R) problem, aiming to achieve a competitive equilibrium that concurrently optimizes seller revenue and buyer utility. We introduce BANTER, an iterative algorithm that jointly determines (1) optimal NFT purchases for buyers (via NFT-REC), considering breeding utility and current prices; and (2) optimal pricing for sellers (via PRICEREC), based on aggregated demand from NFT-REC. To efficiently manage the combinatorial complexity of breeding, we devise Optimal Parent Pair Selection (OPPS) and Heterogeneous Parent Set Selection (HPSS) schemes. Theoretical analysis guarantees BANTER to converge to a competitive equilibrium. Experiments on five real-world NFT datasets demonstrate its effectiveness in enhancing both seller revenue and average buyer utility. Source code: https://github.com/jimmy-academia/BANTER

Few-shot learning (FSL) aims to classify unseen examples (query set) into labeled data (support set) through low-dimensional embeddings. However, the diversity and unpredictability of environments and capture devices make FSL more challenging in real-world applications. In this paper, we propose Dual Support Query Shift (DSQS), a novel challenge in FSL that integrates two key issues: inter-set shifts (between support and query sets) and intra-set shifts (within each set), which significantly hinder model performance. To tackle these challenges, we introduce a Dual Alignment framework (DUAL), whose core insight is that clean features can improve optimal transportation (OT) alignment. Firstly, DUAL leverages a robust embedding function enhanced by a repairer network trained with perturbed and adversarially generated “hard” examples to obtain clean features. Additionally, it incorporates a two-stage OT approach with a negative entropy regularizer, which aligns support set instances, minimizes intra-class distances, and uses query data as anchor nodes to achieve effective distribution alignment. We provide a theoretical bound of DUAL and experimental results on three image datasets, compared against 10 state-of-the-art baselines, showing that DUAL achieves a remarkable average performance improvement of 25.66%. Our code is available at https://github.com/siyang-jiang/DUAL.

Generative document retrieval (GDR) uses pre-trained Transformer-based large language models (LLMs) to extract contextual information and directly predict document identifier token sequences, outperforming traditional document retrieval methods. However, LLMs incur significant computational costs, hindering GDR’s practical application and making inference acceleration essential. Early exiting is one of the conditional computing techniques that expedites LLM inference, but it faces challenges when integrated into GDR due to GDR’s semantically hierarchical structured identifiers, which cause error amplification from premature exits. Moreover, although beam search expands the search space, the hierarchical structure of document identifiers restricts the diversity of initial tokens, leading to inefficiencies. In this work, we introduce Bi-Level Early Exiting for Generative Document Retrieval (BiLEE), comprising Layer Level Early Exiting (LLEE) and Token Level Early Exiting (TLEE). LLEEare designed for hierarchical document identifiers, dynamically escaping from the middle layer of the Transformer calculation based on a data-driven calibrated token threshold. TLEE exiting from unpromising candidate sequences, thus discarding unpromising search beams and enhancing beam search efficiency. Both components dynamically balance the speed-to-accuracy trade-offs for different token positions, doubling GDR’s inference speed and obtaining 13× reduction for FLOPs while maintaining the same level of accuracy. Source code: https://github.com/Rui-Fang/BiLEE.

Reinforcement learning (RL) has achieved impressive performance in various domains. However, most RL frameworks oversimplify the problem by assuming a fixed-yet-known environment and often have difficulty being generalized to real-world scenarios. In this paper, we address a new challenge with a more realistic setting, Incremental Reinforcement Learning, where the search space of the Markov Decision Process continually expands. While previous methods usually suffer from the lack of efficiency in exploring the unseen transitions, especially with increasing search space, we present a new exploration framework named Dual-Adaptive ϵ-greedy Exploration (DAE) to address the challenge of Incremental RL. Specifically, DAE employs a Meta Policy and an Explorer to avoid redundant computation on those sufficiently learned samples. Furthermore, we release a testbed based on a synthetic environment and the Atari benchmark to validate the effectiveness of any exploration algorithms under Incremental RL. Experimental results demonstrate that the proposed framework can efficiently learn the unseen transitions in new environments, leading to notable performance improvement, i.e., an average of more than 80%, over eight baselines examined.

While graph neural networks (GNNs) have achieved notable success in various graph mining tasks, conventional GNNs only model the pairwise correlation in 1-hop neighbors without considering the long-term relations and the high-order patterns, thus limiting their performances. Recently, several works have addressed these issues by exploring the motif, i.e., frequent subgraphs. However, these methods usually require an unacceptable computational time to enumerate all possible combinations of motifs. In this paper, we introduce a new GNN framework, namely Random Walk Conformer (RWC), to exploit global correlations and local patterns based on the random walk, which is a promising method to discover the graph structure. Besides, we propose random walk encoding to help RWC capture topological information, which is proven more expressive than conventional spatial encoding. Extensive experiment results manifest that RWC achieves state-of-the-art performance on graph classification and regression tasks. The source code of RWC is available at https://github.com/b05901024/RandomWalkConformer.

Data poisoning attacks against recommender systems (RecSys) often assume a single seller as the adversary. However, in reality, there are usually multiple sellers attempting to promote their items through RecSys manipulation. To obtain the best data poisoning plan, it is important for an attacker to anticipate and withstand the actions of his opponents. This work studies the problem of Multiplayer Comprehensive Attack (MCA) from the perspective of the attacker, considering the subsequent attacks by his opponents. In MCA, we target the Heterogeneous RecSys, where user-item interaction records, user social network, and item correlation graph are used for recommendations. To tackle MCA, we present the Multilevel Stackelberg Optimization over Progressive Differentiable Surrogate (MSOPDS). The Multilevel Stackelberg Optimization (MSO) method is used to form the optimum strategies by solving the Stackelberg game equilibrium between the attacker and his opponents, while the Progressive Differentiable Surrogate (PDS) addresses technical challenges in deriving gradients for candidate poisoning actions. Experiments on Heterogeneous RecSys trained with public datasets show that MSOPDS outperforms all examined prior works by up to 10.6% in average predicted ratings and up to 11.4% in HitRate@3 for an item targeted by an attacker facing one opponent. Source code provided in https://github.com/jimmy-academia/MSOPDS.

The evaluation of participant contribution in federated learning (FL) has recently gained significant attention due to its applicability in various domains, such as incentive mechanisms, robustness enhancement, and client selection. Previous approaches have predominantly relied on the widely adopted Shapley value for participant evaluation. However, the computation of the Shapley value is expensive, despite using techniques like gradient-based model reconstruction and truncating unnecessary evaluations. Therefore, we present an efficient approach called Single-round Participants Amalgamation for Contribution Evaluation (SPACE). SPACE incorporates two novel components, namely Federated Knowledge Amalgamation and Prototype-based Model Evaluation to reduce the evaluation effort by eliminating the dependence on the size of the validation set and enabling participant evaluation within a single communication round. Experimental results demonstrate that SPACE outperforms state-of-the-art methods in terms of both running time and Pearson’s Correlation Coefficient (PCC). Furthermore, extensive experiments conducted on applications, client reweighting, and client selection highlight the effectiveness of SPACE. The code is available at https://github.com/culiver/SPACE.

The phenomena of influence diffusion on social networks have received tremendous research interests in the past decade. While most prior works mainly focus on predicting the total influence spread on a single network, a marketing campaign that exploits influence diffusion often involves multiple channels with various information disseminated on different media. In this paper, we introduce a new influence estimation problem, namely Content-aware Multi-channel Influence Diffusion (CMID), and accordingly propose CMINet to predict newly influenced users, given a set of seed users with different multimedia contents. In CMINet, we first introduce DiffGNN to encode the influencing power of users (nodes) and Influence-aware Optimal Transport (IOT) to align the embeddings to address the distribution shift across different diffusion channels. Then, we transform CMID into a node classification problem and propose Social-based Multimedia Feature Extractor (SMFE) and Content-aware Multi-channel Influence Propagation (CMIP) to jointly learn the user preferences on multimedia contents and predict the susceptibility of users. Furthermore, we prove that CMINet preserves monotonicity and submodularity, thus enabling (1 − 1/e)-approximate solutions for influence maximization. Experimental results manifest that CMINet outperforms eleven baselines on three public datasets.

Efficient matrix operations have been deemed keys to efficient data analysis. Dual-Triangular QR Decomposition (DT-QRD) is a critical component in Tall and skinny QR decomposition (TS-QRD), which is a widely-used matrix operation with various applications, such as data compression and feature extraction. In order to accelerate DT-QRD, in this paper, we propose a new acceleration framework, including Global Acceleration Schemes, and Partially$\boldsymbol{Q}$-rotation Skipping, which utilize the special DT structure in both$\mathbf{Q}$and$\mathbf{R}$matrix to reduce the latency and computation resource. Further, we employ the Systolic-Array Based Architecture (1D & 2D) for implementation to reduce the memory usage. Experimental results manifest that our framework achieves$169.70\times\ (\mathbf{1}\mathbf{D})$and$250.13\times\ (\mathbf{2}\mathbf{D})$speedup.

Online social networks have become a crucial medium to disseminate the latest political, commercial, and social information. Users with high visibility are often selected as seeds to spread information and affect their adoption in target groups. We study how gender differences and similarities can impact the information spreading process. Using a large-scale Instagram dataset and a small-scale Facebook dataset, we first conduct a multi-faceted analysis taking the interaction type, directionality and frequency into account. To this end, we explore a variety of existing and new single and multihop centrality measures. Our analysis unveils that males and females interact differently depending on the interaction types, e.g., likes or comments, and they feature different support and promotion patterns. We complement prior work showing that females do not reach top visibility (often referred to as the glass ceiling effect) jointly factoring in the connectivity and interaction intensity, both of which were previously mainly discussed independently.

Tall and skinny QR (TSQR) decomposition is an essential matrix operation with various applications in edge computing, including data compression, subspace projection, and dimension reduction. As a critical component in TSQR, Dual-Triangular QR (DTQR) decomposition is solved by the Normal QR method in most works without utilizing the dual-triangular structure. Therefore, we propose a novel DTQR accelerator by recursively exploring the DT structure and propose three acceleration strategies with the systolic array to achieve higher parallelism. Experimental results manifest that our algorithm achieves 21.55x on average speedup compared with the baselines.

Due to the great success of image-to-image (Img2Img) translation GANs, many applications with ethics issues arise, e.g., DeepFake and DeepNude, presenting a challenging problem to prevent the misuse of these techniques. In this work, we tackle the problem by a new adversarial attack scheme, namely the Nullifying Attack, which cancels the image translation process and proposes a corresponding framework, the Limit-Aware Self-Guiding Gradient Sliding Attack (LaS-GSA) under a black-box setting. In other words, by processing the image with the proposed LaS-GSA before publishing, any image translation functions can be nullified, which prevents the images from malicious manipulations. First, we introduce the limit-aware RGF and the gradient sliding mechanism to estimate the gradient that adheres to the adversarial limit, i.e., the pixel value limitations of the adversarial example. We theoretically prove that our model is able to avoid the error caused by the projection in both the direction and the length. Then, an effective self-guiding prior is extracted solely from the threat model and the target image to efficiently leverage the prior information and guide the gradient estimation process. Extensive experiments demonstrate that LaS-GSA requires fewer queries to nullify the image translation process with higher success rates than 4 state-of-the-art methods.

Owing to a wide range of important applications, such as team formation, dense subgraph discovery, and activity attendee suggestions on online social networks, Group Query attracts a lot of attention from the research community. However, most existing works are constrained by a unified social tightness k (e.g., for k-core, or k-plex), without considering the diverse preferences of social cohesiveness in individuals. In this paper, we introduce a new group query, namely Parameter-free Group Query (PGQ), and propose a learning-based model, called PGQN, to find a group that accommodates personalized requirements on social contexts and activity topics. First, PGQN extracts node features by a GNN-based method on Heterogeneous Activity Information Network (HAIN). Then, we transform the PGQ into a graph-to-set (Graph2Set) problem to learn the diverse user preference on topics and members, and find new attendees to the group. Experimental results manifest that our proposed model outperforms nine state-of-the-art methods by at least 51% in terms of F1-score on three public datasets.