GONE: Structural Knowledge Unlearning via Neighborhood-Expanded Distribution Shaping
本文针对大语言模型中知识遗忘问题，提出GONE基准及NEDS框架，有效处理结构化数据中的直接事实移除、推理泄露和灾难性遗忘。
1 citationsRead paper
Selected work
Representative Papers
A Bayesian Discrete Framework for Enhancing Decision-Making Processes in Clinical Trial Designs and Evaluations
This study addresses the limitations of conventional frequentist approaches in effectively incorporating prior knowledge, which constrains adaptive decision-making and reliability in clinical trials. The authors propose a Bayesian framework tailored for discrete probability distributions—such as binomial, Poisson, and negative binomial—to model binary responses and overdispersed clinical endpoints using Bayesian networks. By continuously integrating accumulating evidence, the framework dynamically optimizes trial design and evaluation. Compared to maximum likelihood estimation, this approach demonstrates greater flexibility and robustness in both inferential behavior and practical performance, substantially enhancing decision quality while mitigating misinterpretation of results and reproducibility challenges.
1 citationsRead paper
FoundAna: A GNN-assisted Foundation Model for Graph Anomaly Detection
为解决图异常检测中的迁移性问题，提出FoundAna模型，结合GNN和Transformer捕捉局部与全局结构信息，通过重建误差识别异常。
0 citationsRead paper
DualPathOcc: Dual-Resolution BEV Encoder for 3D Occupancy Prediction
本文提出DualPathOcc框架，通过高分辨率特征聚合和双路径BEV编码器解决多视图图像的3D占用预测问题，优化模型以提高预测精度。
0 citationsRead paper
Membership is Ownership: A Robust Ownership Verification Framework for Diffusion Models
本文提出一种名为MiO的框架，通过成员推断和模型分离验证扩散模型的所有权，以最小化模型效用损失并提高鲁棒性。
0 citationsRead paper
Recent publications
Latest Papers
FoundAna: A GNN-assisted Foundation Model for Graph Anomaly Detection
为解决图异常检测中的迁移性问题，提出FoundAna模型，结合GNN和Transformer捕捉局部与全局结构信息，通过重建误差识别异常。
0 citationsRead paper
DualPathOcc: Dual-Resolution BEV Encoder for 3D Occupancy Prediction
本文提出DualPathOcc框架，通过高分辨率特征聚合和双路径BEV编码器解决多视图图像的3D占用预测问题，优化模型以提高预测精度。
0 citationsRead paper
Membership is Ownership: A Robust Ownership Verification Framework for Diffusion Models
本文提出一种名为MiO的框架，通过成员推断和模型分离验证扩散模型的所有权，以最小化模型效用损失并提高鲁棒性。
0 citationsRead paper
Rethinking the Transferable Adversarial Attacks and Robust Defense in Federated Learning
本文分析了联邦学习中对抗样本的可迁移性，并设计了一种基于对抗训练的防御机制以缓解可迁移对抗样本攻击。
0 citationsRead paper
A Machine Learning Surrogate for Component Criticality Ranking in Interdependent Power-Communication Networks
This study addresses the prohibitive computational cost of high-fidelity simulations in evaluating cascading failures of power-communication coupled systems under large-scale N-k contingencies, which hinders resilience planning. To overcome this challenge, the authors propose a structure-based machine learning surrogate model that, for the first time, integrates leak-free topological centrality measures with cross-layer dependency information to rapidly predict failure severity and generate component criticality rankings. This surrogate model forms the first stage of a two-stage workflow paired with high-fidelity MIIM simulations for prioritized hardening analysis. Evaluated on the IEEE 118-bus system, the model achieves Spearman correlation coefficients of 0.849 and 0.853 for failure severity prediction and criticality ranking, respectively—significantly outperforming purely topological baselines and closely approaching the empirical upper bound of high-fidelity simulation, thereby substantially improving assessment efficiency without compromising accuracy.
0 citationsRead paper