DWDi Wu

Di Wu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Di Wu has a total of 316 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    GRM: Utility-Aware Jailbreak Attacks on Audio LLMs via Gradient-Ratio Masking

    by Yunqiang Wang, Hengyuan Na, Di Wu, Miao Hu, Guocong Quan

  • Open Access

    Compose, Route, Complete: Text-Routed Vision Prototypes for Multimodal Point Cloud Completion

    by Wang Luo, Di Wu, Hengyuan Na, Yinlin Zhu, Shenghao Ye, Miao Hu, Guocong Quan

  • Open Access

    Clinical-Knowledge-Driven Prompt-Tuning for Predicting Occult Lymph Node Metastasis in Early-Stage Head and Neck Cancer: A Multicenter MRI-Based Study

    by Ling Qiu, Guangzi Shi, Zhouwu Liu, Bin Wei, Xaiohua Ban, Qunxing Li, Jin Ye, Xi Zhong, Marco Ventin, Niu Liu, et al.

    Background This retrospective diagnostic study developed and validated an accurate and efficient method to predict occult cervical lymph node metastasis (LNM) for early-stage head and neck cancer (HNC). Method 591 patients were retrospectively recruited from Sun Yat-sen Memorial Hospital (n=308), Sun Yat-sen University Cancer Center (n=104), Affiliated Cancer Hospital and Institute of Guangzhou Medical University (n=94), Shanghai Ninth People’s Hospital (n=42), and The Third Affiliated Hospital of Sun Yat-sen University (n=43) between January, 2012, and June, 2023. A prompt-assisted deep learning model (HNC-Net) was developed using MR images of primary tumor sites, and LNM prediction performance was evaluated using the area under the receiver operating characteristic curve (AUROC), sensitivity (SENS), specificity (SPEC), positive predictive value (PPV), and negative predictive value (NPV). Results HNC-Net had favorable performance for predicting LNM in early-stage HNC patients both in the training dataset (AUROC 0.89 [0.85, 0.93], SENS 0.92 [0.87, 0.98], SPEC 0.66 [0.60, 0.73], PPV 0.54 [0.46, 0.61], NPV 0.95 [0.92, 0.99]) and in the testing datasets (AUROC 0.90 [0.87~0.95], SENS 0.93 [0.87, 0.98], SPEC 0.60 [0.53, 0.67], PPV 0.48 [0.40, 0.56], NPV 0.95 [0.92, 0.99]). The model demonstrated exceptional performance across various subsites (AUROC for oral cavity 0.89 [0.84, 0.94], oropharynx 0.96 [0.86, 0.99], and larynx 0.91 [0.86, 0.99]), and clinical T stages (AUROC for cT1 0.90 [0.72, 0.99], for cT2 0.89 [0.84, 0.94]). Conclusion The MRI-based HNC-Net model is a promising diagnostic tool for predicting cervical LNM in early-stage HNC, which will help surgeons make appropriate surgical decisions regarding neck dissection.

  • A Novel Parallel Approach with Index Optimization for Truss Maintenance

    by Wen Bai, Yufeng Wang, Ke Zheng, Chongwu Dong, Di Wu

    The $k$-truss is a type of cohesive subgraph where each edge is contained in at least $k-2$ triangles within the subgraph, and it is commonly used in community search and dense subgraph discovery. Although $k$-trusses can be computed in polynomial time, obtaining all $k$-trusses of a dynamic graph that evolves over time through edge insertions and deletions remains computationally expensive. To address this challenge, prior studies proposed truss maintenance approaches that update the affected $k$-trusses when the dynamic graph changes. However, existing approaches only utilize one tool of parallelism or auxiliary indexes due to latent conflicts, which restricts their overall performance. To resolve this limitation, we propose a novel parallel approach with index optimization that effectively resolves the conflicts between parallelism and index maintenance. Specifically, we first establish a novel truss maintenance theory based on auxiliary indexes, which ensures the feasibility and correctness of our maintenance solution. Second, we design two effective edge partitioning strategies to improve the parallelism of the proposed algorithms. Finally, we implement two efficient parallel algorithms that cooperate with auxiliary indexes to reduce maintenance costs. Extensive experiments on real-world graphs demonstrate that our methods outperform all baseline approaches by up to one order of magnitude.

  • Open Access

    Both Topology and Text Matter: Revisiting LLM-guided Out-of-Distribution Detection on Text-attributed Graphs

    by Yinlin Zhu, Di Wu, Xu Wang, Guocong Quan, Miao Hu

    Text-attributed graphs (TAGs) associate nodes with textual attributes and graph structure, enabling GNNs to jointly model semantic and structural information. While effective on in-distribution (ID) data, GNNs often encounter out-of-distribution (OOD) nodes with unseen textual or structural patterns in real-world settings, leading to overconfident and erroneous predictions in the absence of reliable OOD detection. Early approaches address this issue from a topology-driven perspective, leveraging neighboring structures to mitigate node-level detection bias. However, these methods typically encode node texts as shallow vector features, failing to fully exploit rich semantic information. In contrast, recent LLM-based approaches generate pseudo OOD priors by leveraging textual knowledge, but they suffer from several limitations: (1) a reliability-informativeness imbalance in the synthesized OOD priors, as the generated OOD exposures either deviate from the true OOD semantics, or introduce non-negligible ID noise, all of which offers limited improvement to detection performance; (2) reliance on specialized architectures, which prevents incorporation of the extensive effective topology-level insights that have been empirically validated in prior work. To this end, we propose LG-Plug, an LLM-Guided Plug-and-play strategy for TAG OOD detection tasks. LG-Plug aligns topology and text representations to produce fine-grained node embeddings, then generates consensus-driven OOD exposure via clustered iterative LLM prompting. Moreover, it leverages lightweight in-cluster codebook and heuristic sampling reduce time cost of LLM querying. The resulting OOD exposure serves as a regularization term to separate ID and OOD nodes, enabling seamless integration with existing detectors.

  • Toward Efficient Wireless Federated Learning via Decoupling Over-the-Air Model Aggregation and Client Selection

    by Saqr Thabet, Yipeng Zhou, Behnaz Soltani, Quan Z. Sheng, Shiting Wen, Di Wu

    Federated learning (FL) is revolutionizing machine learning by enabling multiple decentralized clients to collaboratively train a shared model. In mobile scenarios, client devices exchange model parameters with a central server via wireless channels. However, designing efficient wireless FL (WFL) is challenging due to limited energy and channel capacity. To fully utilize communication resources, over-the-air (OTA) computation has been introduced, allowing direct aggregation of analog parameter signals. However, it conceals clients’ information from the server, making advanced client selection strategies, e.g., cluster-based client selection, and sparsification compression algorithms like TopK inapplicable. To address these limitations, we propose the WFL with Voting-based Clustering (WFL-VC) algorithm, which can integrate advanced client selection and the TopK model sparsification algorithm with OTA computation. WFL-VC consists of two phases: 1) Phase 1: clients vote on significant parameters based on local models, allowing the server to select clients and identify the global Topk parameters; and 2) Phase 2: the selected clients upload model update parameters with globally aligned indices for over-the-air computation at the server. By combining OTA computation with cluster-based client selection and TopK sparsification, WFL-VC substantially reduces the energy consumption of WFL. Extensive experiments on real-world datasets show that WFL-VC outperforms competitive baselines while consuming considerably less energy.

  • Open Access

    Rethinking Federated Graph Foundation Models: A Graph-Language Alignment-based Approach

    by Yinlin Zhu, Di Wu, Xianzhi Zhang, Yuming Ai, Xunkai Li, Miao Hu, Guocong Quan

    Recent studies of federated graph foundational models (FedGFMs) break the idealized and untenable assumption of having centralized data storage to train graph foundation models, and accommodate the reality of distributed, privacy-restricted data silos. Despite their simplicity and intuition, existing studies that project aligned generalizable knowledge onto a discrete token space via vector-quantized backbones suffer from irreversible knowledge loss during the quantization process. In this context, we argue that reconciling the semantic-structural orthogonality and integrity between pre-trained language models (PLMs) and graph neural networks (GNNs) is paramount for developing effective FedGFMs while simultaneously mitigating the severe data heterogeneity and communication constraints inherent in distributed, resource-limited environments. To address these issues, we propose FedGALA (Federated Graph And Language Alignment), a framework that resolves graph-based semantic-structural orthogonality and integrity in federated settings by employing unsupervised contrastive learning to align GNNs and frozen PLMs within a continuous embedding space, thereby capturing robust, transferable general knowledge. Subsequently, FedGALA leverages a communication-efficient prompt tuning mechanism to steer these pre-aligned encoders and frozen PLMs, facilitating effective adaptation to diverse downstream tasks while circumventing the prohibitive overhead of full-parameter fine-tuning. The comprehensive experiments validate that FedGALA outperforms all competitive baselines across multi-domain datasets on multiple tasks with up to 14.37% performance improvement.

  • Fairness-Aware Federated Recommender Design With Heterogeneous Privacy Budgets

    by Miao Hu, Yang Li, Pengshan Liao, Yipeng Zhou, Di Wu

    To mitigate the rising concern on privacy infringement in recommenders, federated recommender (FRec) boosted with local differential privacy (LDP) has been proposed, in which each user privately retains his (or her) dataset, and only exposes model parameters distorted by LDP to a parameter server (PS). Although such kind of solution can largely preserve data privacy, it overlooks the potential unfairness issue. For example, a user can become a free-rider by setting an extremely small privacy budget to completely distort his (or her) model such that the model training is mainly contributed by other users. To guarantee fairness, we design a fairness-aware federated recommender framework called FFRec when users adopt heterogeneous privacy budgets. In FFRec, fairness implies that clients with higher utilities will have more opportunities to participate in training. We integrate fairness into FFRec by taking both sample size and privacy requirements into account. Specifically, a user is selected by the PS to participate in model training based on his (or her) sample size and potential noise influence by LDP and cumulative participation counts. We formulate the fairness-aware client selection problem as an integer programming problem and then reduce it into a submodular problem which has been proven to be NP-hard. To resolve this problem, we employ an approximation algorithm with an approximation ratio of 1 − e−12. Extensive experiments with popular field-measured datasets (i.e., Movielens and Netflix) demonstrate that FFRec can improve fairness by 10.3%-38.5% while guaranteeing high recommendation accuracy