Jiahao Liu (刘家豪)

About Me

Jiahao Liu is a PhD candidate (2021.9-2026.12 expected) in Computer Science at Fudan University. He holds a Bachelor's degree from Southwest Jiaotong University. His research focuses on personalized AI and recommender systems, integrating LLM-driven agents, interactive technologies, causal inference, and graph learning to explore efficient and trustworthy information distribution paradigms.

I am currently on the job market looking for new opportunities. (目前正在就业市场寻找机会)

Work Experience

Publications

(* indicates equal contributions.)

Main Author

• Distribution-Aware End-to-End Embedding for Streaming Numerical Features in Click-Through Rate Prediction
  Jiahao Liu, Hongji Ruan, Weimin Zhang et al.
  KDD 2026 CCF A
  Highlight: Developed an end-to-end distribution-aware embedding framework for streaming numerical features, yielding a +2.307% increase in advertiser value during online A/B testing.
• UniGCRec: Unified User-Item Quantization for Generative Cross-Domain Recommendation
  Chaoyue Ding, Jiahao Liu, Dongsheng Li et al.
  KDD 2026 CCF A
  Highlight: Jointly quantizes users and items into a shared semantic-collaborative ID space to enable preference-aware transfer in low-overlap cross-domain recommendation.
• Drift-Aware Incremental Token Adaptation with Collaborative Semantics for Generative Recommendation
  Yuebo Feng, Jiahao Liu, Mingzhe Han et al.
  SIGIR 2026 CCF A
  Highlight: Proposed a differentiated update and hierarchical code reassignment method to mitigate collaborative representation drift in generative recommender systems, balancing adaptation to new interactions with the stability of token alignment.
• RQ-GMM: Residual Quantized Gaussian Mixture Model for Multimodal Semantic Discretization in CTR Prediction
  Ziye Tong*, Jiahao Liu*, Weimin Zhang et al.
  SIGIR 2026 CCF A
  Highlight: Proposed a residual quantized GMM for multimodal semantic discretization, achieving a significant +1.502% advertiser value improvement in online A/B testing on a short-video platform with hundreds of millions of DAU.
• LLM Agent-based Shilling Attack on Recommender Systems
  Shengkang Gu*, Jiahao Liu*, Dongsheng Li et al.
  WSDM 2026 CCF B
  Highlight: Manipulated recommendation results without accessing internal system data by utilizing LLMs to build fake user agents that generate realistic ratings and reviews.
• EvalAgent: Towards Evaluating News Recommender Systems with LLM-based Agents
  Guangping Zhang, Peng Zhang, Jiahao Liu et al.
  CIKM 2025 CCF B
  Highlight: Introduced a stable memory mechanism to solve noise accumulation in user simulation, providing an accurate, scalable, and low-risk evaluation tool for recommender systems.
• Unbiased Collaborative Filtering with Fair Sampling
  Jiahao Liu, Dongsheng Li, Hansu Gu et al.
  SIGIR 2025 CCF A
  Highlight: Theoretically proved and proposed a simple yet effective sampling strategy to eliminate popularity bias by balancing positive and negative sample probabilities without complex propensity score estimation.
• Improving LLM-powered Recommendations with Personalized Information
  Jiahao Liu*, Xueshuo Yan*, Dongsheng Li et al.
  SIGIR 2025 CCF A
  Highlight: Explicitly extracted user preferences via Chain-of-Thought (CoT), enhancing traditional models with LLM-generated info at the retrieval stage to achieve full-chain performance improvement.
• AgentCF++: Memory-enhanced LLM-based Agents for Popularity-aware Cross-domain Recommendations
  Jiahao Liu*, Shengkang Gu*, Dongsheng Li et al.
  SIGIR 2025 CCF A
  Highlight: Solved cross-domain interference via dual-layer memory and allowed agents to simulate popularity-driven dynamics by grouping interests and sharing memory among "like-minded" users.
• Filtering Discomforting Recommendations with Large Language Models
  Jiahao Liu, Yiyang Shao, Peng Zhang et al.
  WWW 2025 CCF A
  Highlight: Provided a complete technical solution to the algorithmic "black box" discomfort issue, using LLMs to grant users initiative in information consumption without disrupting the existing ecosystem.
• Bidirectional Knowledge Distillation for Enhancing Sequential Recommendation with Large Language Models
  Jiongran Wu*, Jiahao Liu*, Dongsheng Li et al.
  arXiv:2505
  Highlight: Proposed a bidirectional mutual distillation framework to resolve the conflict between the weak semantic understanding of small models and the high inference costs of LLMs.
• Recommendation Unlearning via Matrix Correction
  Jiahao Liu, Dongsheng Li, Hansu Gu et al.
  arXiv:2307
  Highlight: Achieved efficient unlearning by decoupling users' sensitive personal interaction data from item similarities (model parameters).
• AutoSeqRec: Autoencoder for Efficient Sequential Recommendation
  Sijia Liu*, Jiahao Liu*, Hansu Gu et al.
  CIKM 2023 CCF B
  Highlight: Introduced item transition information to propose the first autoencoder-based sequential recommendation model.
• Triple Structural Information Modelling for Accurate, Explainable and Interactive Recommendation
  Jiahao Liu, Dongsheng Li, Hansu Gu et al.
  SIGIR 2023 CCF A
  Highlight: Proposed an incremental SVD technique for the efficient global optimization of embedding representations.
• Parameter-free Dynamic Graph Embedding for Link Prediction
  Jiahao Liu, Dongsheng Li, Hansu Gu et al.
  NeurIPS 2022 CCF A
  Highlight: Proposed a Graph-Transformer based architecture for sequential recommendation and introduced the concept of "Generalized Collaborative Filtering".
• Personalized Graph Signal Processing for Collaborative Filtering
  Jiahao Liu, Dongsheng Li, Hansu Gu et al.
  WWW 2023 CCF A
  Highlight: Interpreted recall methods from a frequency domain perspective and proposed concepts of "personalized graph signals", "augmented similarity graphs", and "mixed frequency filters".

Co-Author

• Feature-Indexed Federated Recommendation with Residual-Quantized Codebooks
  Mingzhe Han, Jiahao Liu, Dongsheng Li, Hansu Gu et al.
  arXiv:2601
• FedCIA: Federated Collaborative Information Aggregation for Privacy-Preserving Recommendation
  Mingzhe Han, Dongsheng Li, Jiafeng Xia, Jiahao Liu et al.
  SIGIR 2025 CCF A
• FIRE: Fast Incremental Recommendation with Graph Signal Processing
  Jiafeng Xia, Dongsheng Li, Hansu Gu, Jiahao Liu et al.
  WWW 2022 CCF A

Honors & Awards

• Doctoral National Scholarship (博士生国家奖学金), 2025
• Doctoral National Scholarship (博士生国家奖学金), 2023
• Undergraduate National Scholarship (本科生国家奖学金), 2020
• Si-Shi-Yang-Hua Gold Medal, Southwest Jiaotong University (竢实扬华奖章，全校学生最高荣誉), 2021
• Outstanding Graduate, Southwest Jiaotong University (优秀毕业生), 2021