Jun Fan
Jun Fan has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Qian Gao
Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan), Qilu University of Technology (Shandong Academy of Sciences), Jinan, China
8 shared publications - Guohui Ding
Shenyang Aerospace University, Shenyang, China
2 shared publications - Peng Xue
Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan), Qilu University of Technology (Shandong Academy of Sciences), Jinan, China
2 shared publications - Zhixue Zhang
Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan) Qilu University of Technology (Shandong Academy of Sciences) Jinan China
2 shared publications - Zongyang Wang
Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan) Qilu University of Technology (Shandong Academy of Sciences) Jinan China
2 shared publications - Nana Wang
Qilu University of Technology (Shandong Academy of Sciences)
1 shared publication
Latest Publications
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 5498-5509
SADI: Spatial-Aware Dynamic Intervention for Mitigating Hallucinations in LVLMs
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 5498-5509
Open AccessSADI: Spatial-Aware Dynamic Intervention for Mitigating Hallucinations in LVLMs
- 30 April 2026
- Published by Wiley in ETRI Journal
LLMSQRec: A study on a large language model‐based framework for dual‐view semantic‐quantized recommendation
Sequential recommendation is widely used in domains such as e‐commerce and social media but still suffers from data sparsity, long‐tail distributions, and popularity bias. Large language models (LLMs) provide rich semantics and transferability, yet their integration with collaborative signals remains challenging due to representation inconsistency, high prompting costs, weakened user modeling, and unresolved bias. To address these issues, we propose LLMSQRec, a unified framework that fuses LLM‐derived semantic features with ID embeddings via dual‐view modeling. Specifically, representation inconsistency is alleviated through semantic‐collaborative alignment; prompting costs are reduced by combining frozen LLM embeddings with lightweight vector quantization; user modeling is strengthened by jointly encoding semantic and collaborative signals, and popularity bias is mitigated through curriculum learning with popularity‐aware regularization. Experiments on multiple real‐world datasets demonstrate that LLMSQRec significantly outperforms existing methods in sparse and long‐tail settings, achieving superior accuracy, robustness, and diversity.
- 30 April 2026
- Published by Wiley in ETRI Journal
LLMSQRec: A study on a large language model‐based framework for dual‐view semantic‐quantized recommendation
Sequential recommendation is widely used in domains such as e‐commerce and social media but still suffers from data sparsity, long‐tail distributions, and popularity bias. Large language models (LLMs) provide rich semantics and transferability, yet their integration with collaborative signals remains challenging due to representation inconsistency, high prompting costs, weakened user modeling, and unresolved bias. To address these issues, we propose LLMSQRec, a unified framework that fuses LLM‐derived semantic features with ID embeddings via dual‐view modeling. Specifically, representation inconsistency is alleviated through semantic‐collaborative alignment; prompting costs are reduced by combining frozen LLM embeddings with lightweight vector quantization; user modeling is strengthened by jointly encoding semantic and collaborative signals, and popularity bias is mitigated through curriculum learning with popularity‐aware regularization. Experiments on multiple real‐world datasets demonstrate that LLMSQRec significantly outperforms existing methods in sparse and long‐tail settings, achieving superior accuracy, robustness, and diversity.
- 23 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. PP (5) , 1-12
SemAlignRec: A Large Language Model-Driven Cross-Modal Collaborative–Semantic Recommendation Framework
Recommender systems are increasingly vital in e-commerce and social media, yet traditional collaborative filtering and deep learning methods still face challenges in fine-grained interest modeling and data sparsity. Large language models (LLMs), with strong semantic understanding and reasoning capabilities, excel in cold-start and long-tail scenarios. However, the modality gap between low-dimensional collaborative signals and LLMs’ contextual embeddings leads to fusion difficulties and semantic distortion. To address this, we propose SemAlignRec, a unified recommendation framework that enhances fusion through a model-agnostic collaborative–semantic alignment mechanism. Specifically: 1) deep semantic encoding and graph structural enhancement: leveraging LLMs to encode user and item textual attributes into high-quality semantic embeddings, which are further refined by a multilayer deep semantic encoder; a graph enhancement module incorporating residual connections, node attention, and graph perturbations strengthens collaborative embeddings and facilitates their fusion with semantic embeddings; and 2) transmodal fusion and self-supervised optimization: integrating multihead attention, gated fusion, and contrastive self-supervision to achieve precise fusion of collaborative and semantic embeddings. Experiments on multiple datasets show that SemAlignRec delivers substantial performance gains, achieving an overall average improvement of about 6.6% over the compared baselines. In particular, compared with the latest baseline DALR, SemAlignRec obtains an average gain of approximately 9.8% on key metrics for the Amazon and Yelp datasets, demonstrating its strong practical potential.
- 23 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. PP (5) , 1-12
SemAlignRec: A Large Language Model-Driven Cross-Modal Collaborative–Semantic Recommendation Framework
Recommender systems are increasingly vital in e-commerce and social media, yet traditional collaborative filtering and deep learning methods still face challenges in fine-grained interest modeling and data sparsity. Large language models (LLMs), with strong semantic understanding and reasoning capabilities, excel in cold-start and long-tail scenarios. However, the modality gap between low-dimensional collaborative signals and LLMs’ contextual embeddings leads to fusion difficulties and semantic distortion. To address this, we propose SemAlignRec, a unified recommendation framework that enhances fusion through a model-agnostic collaborative–semantic alignment mechanism. Specifically: 1) deep semantic encoding and graph structural enhancement: leveraging LLMs to encode user and item textual attributes into high-quality semantic embeddings, which are further refined by a multilayer deep semantic encoder; a graph enhancement module incorporating residual connections, node attention, and graph perturbations strengthens collaborative embeddings and facilitates their fusion with semantic embeddings; and 2) transmodal fusion and self-supervised optimization: integrating multihead attention, gated fusion, and contrastive self-supervision to achieve precise fusion of collaborative and semantic embeddings. Experiments on multiple datasets show that SemAlignRec delivers substantial performance gains, achieving an overall average improvement of about 6.6% over the compared baselines. In particular, compared with the latest baseline DALR, SemAlignRec obtains an average gain of approximately 9.8% on key metrics for the Amazon and Yelp datasets, demonstrating its strong practical potential.
- 24 June 2025
- Published by Springer Nature
- p. 169-182
ADERec: Adaptive Data Augmentation Sequence Recommendation Based on Dual Network Architecture
- 24 June 2025
- Published by Springer Nature
- p. 169-182
ADERec: Adaptive Data Augmentation Sequence Recommendation Based on Dual Network Architecture
- 19 April 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (6) , 1-27
Social recommendation based on contrastive learning of hypergraph convolution
- 19 April 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (6) , 1-27
Social recommendation based on contrastive learning of hypergraph convolution
- 26 November 2023
- Published by Springer Nature
- p. 493-506
KSHFS: Research on Drug-Drug Interaction Prediction Based on Knowledge Subgraph and High-Order Feature-Aware Structure
- 26 November 2023
- Published by Springer Nature
- p. 493-506
KSHFS: Research on Drug-Drug Interaction Prediction Based on Knowledge Subgraph and High-Order Feature-Aware Structure
- 16 November 2023
- Published by Springer Nature
- p. 357-369
EWMIGCN: Emotional Weighting Based Multimodal Interaction Graph Convolutional Networks for Personalized Prediction
- 16 November 2023
- Published by Springer Nature
- p. 357-369
EWMIGCN: Emotional Weighting Based Multimodal Interaction Graph Convolutional Networks for Personalized Prediction
- Open Access
- 26 September 2023
- Published by Springer Nature in Research Square
Social Recommendation Based on Contrastive Learning of Hypergraph Convolution
Recommendation systems often use social relationships to improve recommendation quality in the face of sparse data. However, in reality, user interactions can be very complex, user relationships may be high-order, and the contribution of each neighbor may be different. Therefore, traditional social relationship methods cannot fully explore user interactions. Additionally, data sparsity can lead to poor model robustness, making it susceptible to noisy data. To address these issues, this paper proposes a multi-channel hypergraph convolutional network model based on contrastive learning, aiming to enhance social recommendation through high-order user relationships and contrastive learning. The model includes an embedding layer, a propagation layer, and a contrastive learning layer. For user embedding, each channel of the propagation layer learns high-order embedding information through hypergraph convolution. For item embedding, the neighbor-aware attention coefficient is used to mine the implicit correlations of items. In addition, the contrastive learning layer adds randomly uniform noise to the embedding to perform graph augmentation at the representation level, using contrastive learning to obtain more uniformly distributed embedding representations for score prediction. Experimental results on two real datasets, Douban and Yelp, show that the proposed model improves Recall and NDCG by 1.3%-2.1% and 2.0%-2.8%, respectively.
- 26 September 2023
- Published by Springer Nature in Research Square
Open AccessSocial Recommendation Based on Contrastive Learning of Hypergraph Convolution
Recommendation systems often use social relationships to improve recommendation quality in the face of sparse data. However, in reality, user interactions can be very complex, user relationships may be high-order, and the contribution of each neighbor may be different. Therefore, traditional social relationship methods cannot fully explore user interactions. Additionally, data sparsity can lead to poor model robustness, making it susceptible to noisy data. To address these issues, this paper proposes a multi-channel hypergraph convolutional network model based on contrastive learning, aiming to enhance social recommendation through high-order user relationships and contrastive learning. The model includes an embedding layer, a propagation layer, and a contrastive learning layer. For user embedding, each channel of the propagation layer learns high-order embedding information through hypergraph convolution. For item embedding, the neighbor-aware attention coefficient is used to mine the implicit correlations of items. In addition, the contrastive learning layer adds randomly uniform noise to the embedding to perform graph augmentation at the representation level, using contrastive learning to obtain more uniformly distributed embedding representations for score prediction. Experimental results on two real datasets, Douban and Yelp, show that the proposed model improves Recall and NDCG by 1.3%-2.1% and 2.0%-2.8%, respectively.