Zihao Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zihao Li has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Guandong Xu
The Education University of Hong Kong
12 shared publications - Xianzhi Wang
University of Technology Sydney, Sydney, Australia
12 shared publications - Yakun Chen
Center for Learning, Teaching and Technology, The Education University of Hong Kong, Hong Kong, 999077, Hong Kong, China
11 shared publications - Chao Yang
Faculty of Information Science and Engineering, Ocean University of China, Qingdao, China
8 shared publications - Lina Yao
University of New South Wales, Sydney, Australia
5 shared publications - Tong Zhang
University of Technology Sydney, Sydney, Australia
4 shared publications
Latest Publications
- 12 May 2026
- Published by Elsevier in Neural Networks
- Vol. 202, 109088
WEANet: Bridging wavelet inductive bias with network parameter initialization for time series modeling
- 12 May 2026
- Published by Elsevier in Neural Networks
- Vol. 202, 109088
WEANet: Bridging wavelet inductive bias with network parameter initialization for time series modeling
- Open Access
- 5 March 2026
- Published by Elsevier in SSRN Electronic Journal
NeighCF: Neighborhood-aware Collaborative Filtering for LLM-based Sequential Recommendation
Recently, Large Language Models (LLMs) have shown strong promise in recommendation tasks by leveraging semantic reasoning and external knowledge. As inputs to LLMs, prompts play a crucial role in guiding LLMs to generate outputs that align with task-specific requirements. However, existing prompting strategies primarily rely on a user’s individual history, overlooking collaborative signals that capture broader behavioral patterns. While recent works attempt to incorporate neighbors’ histories, they face two key challenges: (i) fixed-size neighbor selection, which either underutilizes sparse contexts or overloads dense ones with redundancy; and (ii) computationally expensive global neighbor retrieval, which is impractical for large-scale systems. Specifically, NeighCF (i) employs a cluster-aware retrieval strategy that restricts neighbor search to candidate-relevant subspaces, reducing the cost of global retrieval while maintaining high recall; and (ii) introduces an adaptive neighbor selection mechanism based on a learnable threshold that dynamically adjusts the number of neighbors according to interaction density and candidate relevance. The resulting neighbor representations are integrated into prompts for LLM-based reasoning. Extensive experiments on various datasets demonstrate that NeighCF consistently outperforms both single-user prompting baselines and state-of-the-art collaborative prompting approaches. These results validate the effectiveness and efficiency of our framework, highlighting collaborative reasoning in LLM-based sequential recommendation.
- 5 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessNeighCF: Neighborhood-aware Collaborative Filtering for LLM-based Sequential Recommendation
Recently, Large Language Models (LLMs) have shown strong promise in recommendation tasks by leveraging semantic reasoning and external knowledge. As inputs to LLMs, prompts play a crucial role in guiding LLMs to generate outputs that align with task-specific requirements. However, existing prompting strategies primarily rely on a user’s individual history, overlooking collaborative signals that capture broader behavioral patterns. While recent works attempt to incorporate neighbors’ histories, they face two key challenges: (i) fixed-size neighbor selection, which either underutilizes sparse contexts or overloads dense ones with redundancy; and (ii) computationally expensive global neighbor retrieval, which is impractical for large-scale systems. Specifically, NeighCF (i) employs a cluster-aware retrieval strategy that restricts neighbor search to candidate-relevant subspaces, reducing the cost of global retrieval while maintaining high recall; and (ii) introduces an adaptive neighbor selection mechanism based on a learnable threshold that dynamically adjusts the number of neighbors according to interaction density and candidate relevance. The resulting neighbor representations are integrated into prompts for LLM-based reasoning. Extensive experiments on various datasets demonstrate that NeighCF consistently outperforms both single-user prompting baselines and state-of-the-art collaborative prompting approaches. These results validate the effectiveness and efficiency of our framework, highlighting collaborative reasoning in LLM-based sequential recommendation.
- 26 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. PP, 1-14
Listwise Preference Alignment Optimization for Tail Item Recommendation
Preference alignment has achieved greater success on large language models (LLMs) and drawn broad interest in recommendation research. Existing preference alignment methods for recommendation either require explicit reward modeling or only support pairwise preference comparison. The former directly increases substantial computational costs, while the latter hinders training efficiency on negative samples. Moreover, no existing effort has explored preference alignment solutions for tail-item recommendation. To bridge the above gaps, we propose LPO4Rec, which extends the Bradley-Terry model from pairwise comparison to listwise comparison, to improve the efficiency of model training. Specifically, we derive a closed-form optimal policy to enable more efficient and effective training without explicit reward modeling. We also present an adaptive negative sampling and reweighting strategy to prioritize tail items during optimization and enhance performance in tail-item recommendations. Besides, we theoretically prove that optimizing the listwise preference optimization (LPO) loss is equivalent to maximizing the upper bound of the optimal reward. Our experiments on three public datasets show that our method outperforms 10 baselines by a large margin, achieving up to 50% performance improvement while reducing 17.9% GPU memory usage when compared with direct preference optimization (DPO) in tail-item recommendation.
- 26 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. PP, 1-14
Listwise Preference Alignment Optimization for Tail Item Recommendation
Preference alignment has achieved greater success on large language models (LLMs) and drawn broad interest in recommendation research. Existing preference alignment methods for recommendation either require explicit reward modeling or only support pairwise preference comparison. The former directly increases substantial computational costs, while the latter hinders training efficiency on negative samples. Moreover, no existing effort has explored preference alignment solutions for tail-item recommendation. To bridge the above gaps, we propose LPO4Rec, which extends the Bradley-Terry model from pairwise comparison to listwise comparison, to improve the efficiency of model training. Specifically, we derive a closed-form optimal policy to enable more efficient and effective training without explicit reward modeling. We also present an adaptive negative sampling and reweighting strategy to prioritize tail items during optimization and enhance performance in tail-item recommendations. Besides, we theoretically prove that optimizing the listwise preference optimization (LPO) loss is equivalent to maximizing the upper bound of the optimal reward. Our experiments on three public datasets show that our method outperforms 10 baselines by a large margin, achieving up to 50% performance improvement while reducing 17.9% GPU memory usage when compared with direct preference optimization (DPO) in tail-item recommendation.
- Open Access
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 4181-4190
SarRec: Statistically-guaranteed Augmented Retrieval for Recommendation
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 4181-4190
Open AccessSarRec: Statistically-guaranteed Augmented Retrieval for Recommendation
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 9969-9977
Bridging the Gap: Consistent Image Outpainting via Training-Free Noise Optimization
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 9969-9977
Open AccessBridging the Gap: Consistent Image Outpainting via Training-Free Noise Optimization
- Open Access
Listwise Preference Alignment Optimization for Tail Item Recommendation
Preference alignment has achieved greater success on Large Language Models (LLMs) and drawn broad interest in recommendation research. Existing preference alignment methods for recommendation either require explicit reward modeling or only support pairwise preference comparison. The former directly increases substantial computational costs, while the latter hinders training efficiency on negative samples. Moreover, no existing effort has explored preference alignment solutions for tail-item recommendation. To bridge the above gaps, we propose LPO4Rec, which extends the Bradley-Terry model from pairwise comparison to listwise comparison, to improve the efficiency of model training. Specifically, we derive a closed form optimal policy to enable more efficient and effective training without explicit reward modeling. We also present an adaptive negative sampling and reweighting strategy to prioritize tail items during optimization and enhance performance in tail-item recommendations. Besides, we theoretically prove that optimizing the listwise preference optimization (LPO) loss is equivalent to maximizing the upper bound of the optimal reward. Our experiments on three public datasets show that our method outperforms 10 baselines by a large margin, achieving up to 50% performance improvement while reducing 17.9% GPU memory usage when compared with direct preference optimization (DPO) in tail-item recommendation. Our code is available at https://github.com/Yuhanleeee/LPO4Rec.
Open AccessListwise Preference Alignment Optimization for Tail Item Recommendation
Preference alignment has achieved greater success on Large Language Models (LLMs) and drawn broad interest in recommendation research. Existing preference alignment methods for recommendation either require explicit reward modeling or only support pairwise preference comparison. The former directly increases substantial computational costs, while the latter hinders training efficiency on negative samples. Moreover, no existing effort has explored preference alignment solutions for tail-item recommendation. To bridge the above gaps, we propose LPO4Rec, which extends the Bradley-Terry model from pairwise comparison to listwise comparison, to improve the efficiency of model training. Specifically, we derive a closed form optimal policy to enable more efficient and effective training without explicit reward modeling. We also present an adaptive negative sampling and reweighting strategy to prioritize tail items during optimization and enhance performance in tail-item recommendations. Besides, we theoretically prove that optimizing the listwise preference optimization (LPO) loss is equivalent to maximizing the upper bound of the optimal reward. Our experiments on three public datasets show that our method outperforms 10 baselines by a large margin, achieving up to 50% performance improvement while reducing 17.9% GPU memory usage when compared with direct preference optimization (DPO) in tail-item recommendation. Our code is available at https://github.com/Yuhanleeee/LPO4Rec.
- Open Access
- 22 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 4925-4936
Reembedding and Reweighting are Needed for Tail Item Sequential Recommendation
- 22 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 4925-4936
Open AccessReembedding and Reweighting are Needed for Tail Item Sequential Recommendation
- 15 April 2025
- Published by Springer Nature in International Journal of Multimedia Information Retrieval
- Vol. 14 (2) , 1-18
Deep multimodal learning for time series analysis in social computing: a survey
- 15 April 2025
- Published by Springer Nature in International Journal of Multimedia Information Retrieval
- Vol. 14 (2) , 1-18
Deep multimodal learning for time series analysis in social computing: a survey
- Open Access
- 7 November 2024
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 57 (2) , 1-37
Graph and Sequential Neural Networks in Session-based Recommendation: A Survey
Recent years have witnessed the remarkable success of recommendation systems (RSs) in alleviating the information overload problem. As a new paradigm of RSs, session-based recommendation (SR) specializes in users’ short-term preferences and aims at providing a more dynamic and timely recommendation based on ongoing interactions. This survey presents a comprehensive overview of the recent works on SR. First, we clarify the key definitions within SR and compare the characteristics of SR against other recommendation tasks. Then, we summarize the existing methods in two categories: sequential neural network based methods and graph neural network (GNN) based methods. The relevant frameworks and technical details are further introduced. Finally, we discuss the challenges of SR and new research directions in this area.
- 7 November 2024
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 57 (2) , 1-37
Open AccessGraph and Sequential Neural Networks in Session-based Recommendation: A Survey
Recent years have witnessed the remarkable success of recommendation systems (RSs) in alleviating the information overload problem. As a new paradigm of RSs, session-based recommendation (SR) specializes in users’ short-term preferences and aims at providing a more dynamic and timely recommendation based on ongoing interactions. This survey presents a comprehensive overview of the recent works on SR. First, we clarify the key definitions within SR and compare the characteristics of SR against other recommendation tasks. Then, we summarize the existing methods in two categories: sequential neural network based methods and graph neural network (GNN) based methods. The relevant frameworks and technical details are further introduced. Finally, we discuss the challenges of SR and new research directions in this area.
- Open Access
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 2240-2249
TEXT CAN BE FAIR: Mitigating Popularity Bias with PLMs by Learning Relative Preference
Recently, the item textual information has been exploited with pre-trained language models (PLMs) to enrich the representations of tail items. The underlying idea is to align the hot items and tail items in terms of the external semantic knowledge covered by the PLM. However, it is non-trivial to eliminate the popularity bias by exploiting the textual semantics. One major obstacle is that the model supervision still counts on the sparse yet binary user behaviors. In the preliminary investigation, we discover that text-based recommendations also suffer from the popularity bias. To this end, we propose a novel self-distillation framework based on a pre-trained language model, named Staple. The proposed Staple consists of two main components: ranker model and recommender model, which are both instantiated as a PLM towards exploiting the item textual semantics. Motivated by the recent success of reinforcement learning with human feedback (RLHF), the proposed Staple aims to recover the relative preference by learning a fair ranker model that can successfully distinguish the preference levels for uninteracted items. Specifically, analogous to the training of large language models (LLMs), we introduce a pre-training and a fair supervised fine-tuning with a decoupled layer to build the ranker model. Then, similar to RLHF for LLM training, we utilize the relative preference information estimated by the ranker over candidate items to complement the learning of the recommender model. We show that this RLHF process can be reformed as an efficient distillation learning process. We conduct extensive experiments on three real-world datasets. In addition to the performance metrics, we employ two additional metrics to measure fairness and debiased performance. The experiments show that our method can significantly improve the item exposure fairness of recommendation and mitigate popularity bias, while also improving the recommendation performance. The source code is available at https://github.com/WHUIR/STAPLE.
- 21 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 2240-2249
Open AccessTEXT CAN BE FAIR: Mitigating Popularity Bias with PLMs by Learning Relative Preference
Recently, the item textual information has been exploited with pre-trained language models (PLMs) to enrich the representations of tail items. The underlying idea is to align the hot items and tail items in terms of the external semantic knowledge covered by the PLM. However, it is non-trivial to eliminate the popularity bias by exploiting the textual semantics. One major obstacle is that the model supervision still counts on the sparse yet binary user behaviors. In the preliminary investigation, we discover that text-based recommendations also suffer from the popularity bias. To this end, we propose a novel self-distillation framework based on a pre-trained language model, named Staple. The proposed Staple consists of two main components: ranker model and recommender model, which are both instantiated as a PLM towards exploiting the item textual semantics. Motivated by the recent success of reinforcement learning with human feedback (RLHF), the proposed Staple aims to recover the relative preference by learning a fair ranker model that can successfully distinguish the preference levels for uninteracted items. Specifically, analogous to the training of large language models (LLMs), we introduce a pre-training and a fair supervised fine-tuning with a decoupled layer to build the ranker model. Then, similar to RLHF for LLM training, we utilize the relative preference information estimated by the ranker over candidate items to complement the learning of the recommender model. We show that this RLHF process can be reformed as an efficient distillation learning process. We conduct extensive experiments on three real-world datasets. In addition to the performance metrics, we employ two additional metrics to measure fairness and debiased performance. The experiments show that our method can significantly improve the item exposure fairness of recommendation and mitigate popularity bias, while also improving the recommendation performance. The source code is available at https://github.com/WHUIR/STAPLE.