👋About Me关于我

Controllable Generative Recommendation via Guided Token Refinement

Haichao Zhang, Zhixuan Liang, Chong Zhang, Zixi Chen, Wen Wang, Jia Wang

2026 IEEE International Conference on Data Mining (ICDM), accepted

A controllable generative recommendation framework that refines semantic tokens under explicit guidance for more reliable alignment with user intent.一种可控生成式推荐框架，通过显式引导细化语义 token，使生成结果与用户意图更加可靠地对齐。

Explain-then-Forget: Causal Explanation-based Unlearning for Efficient and Precise Recommendation

Haichao Zhang, Chong Zhang, Wen Wang, Shi Qiu, Peiyu Hu, Jia Wang

2026 IEEE International Conference on Data Mining (ICDM), accepted

An efficient recommendation unlearning framework that uses causal explanations to localize and precisely remove the influence of target interactions.一种高效推荐遗忘框架，利用因果解释定位并精确移除目标交互产生的影响。

Explaining Graph Recommendations via Counterfactual Support Sets

Haichao Zhang, Chong Zhang, Can Wang, Shi Qiu, Jia Wang

33rd International Conference on Neural Information Processing (ICONIP 2026), accepted

A counterfactual explanation framework that identifies compact support sets whose removal changes a target user's recommendation without retraining the graph recommender.一种反事实图推荐解释框架，在无需重训练图推荐模型的情况下，识别能够改变目标用户推荐结果的紧凑支持集。

Topology-Aware Inverse Preference Learning for Robust Hybrid Voting Systems

Can Wang, Shi Qiu, Haichao Zhang^*

* Corresponding author* 通讯作者

33rd International Conference on Neural Information Processing (ICONIP 2026), accepted

A topology-aware inverse preference learning framework for reconstructing latent crowd preferences and auditing robustness, conflict, and merit alignment in hybrid voting systems.一种面向混合投票系统的拓扑感知逆偏好学习框架，用于重建潜在人群偏好，并分析鲁棒性、拓扑冲突与能力对齐。