Publication

Embracing Plasticity: Balancing Stability and Plasticity in Continual Recommender Systems [link]

Hyunsik Yoo, SeongKu Kang, Ruizhong Qiu, Charlie Xu, Fei Wang and Hanghang Tong

ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 2025

• We propose PISA, a continual learning framework that adaptively balances stability and plasticity based on user preference shifts.

Personalized Preference Reasoning with Large Language Models for Accurate and Explainable Recommendation [link]

Jieyong Kim, Hyunseo Kim, Hyunjin Cho, SeongKu Kang, Buru Chang, Jinyoung Yeo and Dongha Lee

ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 2025

• We propose an LLM-based recommender optimized via distillation for preference extraction, profile construction, and reasoning-enhanced rating prediction.

Uncertainty Quantification and Decomposition for LLM-based Recommendation [link]

Wonbin Kweon, Sanghwan Jang, SeongKu Kang†, Hwanjo Yu†

ACM The Web Conference (WWW), 2025

• We investigate a systematic approach to quantify and decompose the uncertainty in LLM-based recommendation.

Chain-of-Factors Paper-Reviewer Matching [link]

Yu Zhang, Yanzhen Shen, SeongKu Kang, Xiusi Chen, Bowen Jin, Jiawei Han 

ACM The Web Conference (WWW), 2025

• We propose a unified model for paper-reviewer matching that jointly considers semantic, topic, and citation factors.

Improving Scientific Document Retrieval with Concept Coverage-based Query Set Generation [link]

SeongKu Kang, Bowen Jin, Wonbin Kweon, Yu Zhang, Dongha Lee, Jiawei Han, Hwanjo Yu 

ACM International Conference on Web Search and Data Mining (WSDM), 2025

• We propose CCQGen framework that generates queries with comprehensive coverage of a document's concepts.

Unsupervised Robust Cross-Lingual Entity Alignment via Neighbor Triple Matching with Entity and Relation Texts [link]

Soojin Yoon, Sungho Ko, Tongyoung Kim, SeongKu Kang, Jinyoung Yeo, Dongha Lee 

ACM International Conference on Web Search and Data Mining (WSDM), 2025

• We propose ERAlign, an unsupervised and robust cross-lingual entity alignment pipeline. 

Taxonomy-guided Semantic Indexing for Academic Paper Search [link]

SeongKu Kang, Yunyi Zhang, Pengcheng Jiang, Dongha Lee, Jiawei Han, Hwanjo Yu 

Conference on Empirical Methods in Natural Language Processing (EMNLP), Main (oral), 2024

• We propose Taxonomy-guided Semantic Indexing for effective academic concept matching in paper search.

Continual Collaborative Distillation for Recommender System [link]

Gyuseok Lee*, SeongKu Kang*, Wonbin Kweon, Hwanjo Yu 

ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2024

• We introduce a new research direction that combines knowledge distillation and continual learning for practical recommender systems.

Pearl: A Review-driven Persona-Knowledge Grounded Conversational Recommendation Dataset [link]

Minjin Kim*, Minju Kim*, Hana Kim, Beong-woo Kwak, Soyeon Chun, Hyunseo Kim, SeongKu Kang, Youngjae Yu, Jinyoung Yeo, Dongha Lee

Annual Meeting of the Association for Computational Linguistics (ACL), Findings, 2024

• We present a new conversational recommendation dataset named PEARL, synthesized with persona- and knowledge-augmented LLM simulators.

Self-Consistent Reasoning-based Aspect-Sentiment Quad Prediction with Extract-Then-Assign Strategy [link]

Jieyong Kim*, Ryang Heo*, Yongsik Seo, SeongKu Kang, Jinyoung Yeo, Dongha Lee

Annual Meeting of the Association for Computational Linguistics (ACL), Short paper, Findings, 2024

• We propose SCRAP which generates reasonings and the corresponding sentiment quadruplets in sequence.

Multi-Domain Sequential Recommendation via Domain Space Learning [link]

Junyoung Hwang, Hyunjun Ju, SeongKu Kang, Sanghwan Jang, Hwanjo Yu

ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 2024

• We introduce a new multi-domain sequential recommendation method, specifically targeting the challenging scenario where recent interactions are highly sparse.

Learning Discriminative Dynamics with Label Corruption for Noisy Label Detection [link]

Suyeon Kim, Dongha Lee, SeongKu Kang, Sukang Chae, Sanghwan Jang, Hwanjo Yu

Conference on Computer Vision and Pattern Recognition (CVPR), 2024

• We propose DynaCor that distinguishes mislabeled instances based on the dynamics of the training signals.

Unbiased, Effective, and Efficient Distillation from Heterogeneous Models for Recommender Systems [link]

SeongKu Kang, Wonbin Kweon, Dongha Lee, Jianxun Lian, Xing Xie, Hwanjo Yu

ACM Transactions on Recommender Systems (TORS), 2024

• We leverage dissensus of models to mitigate the popularity amplifications of a large-scale RS.

Improving Retrieval in Theme-specific Applications using a Corpus Topical Taxonomy [link]

SeongKu Kang, Shivam Agarwal, Bowen Jin, Dongha Lee, Hwanjo Yu, Jiawei Han

ACM The Web Conference (WWW), 2024

• We introduce a plug-and-play ToTER framework which improves PLM-based retrieval using a corpus topical taxonomy.

Top-Personalized-K Recommendation [link]

Wonbin Kweon, SeongKu Kang, Sanghwan Jang, Hwanjo Yu

ACM The Web Conference (WWW), 2024

• We propose Top-Personalized-K Recommendation, a new recommendation task aimed at generating a personalized-sized ranking list to maximize individual user satisfaction.

Multi-Domain Recommendation to Attract Users via Domain Preference Modeling [link]

Hyunjun Ju, SeongKu Kang, Dongha Lee, Junyoung Hwang, Sanghwan Jang, Hwanjo Yu

AAAI Conference on Artificial Intelligence (AAAI), 2024

• We propose a new multi-domain recommendation framework that learns various seen-unseen domain mappings in a unified way with masked domain modeling.

MvFS: Multi-view Feature Selection for Recommender System [link]

Youngjune Lee, Yeongjong Jeong, Keunchan Park, SeongKu Kang†

ACM International Conference on Information and Knowledge Management (CIKM), Short paper, 2023

• We propose MvFS, which promotes more balanced feature selection while mitigating bias toward dominant patterns.

Distillation from Heterogeneous Models for Top-K Recommendation [link]

SeongKu Kang, Wonbin Kweon, Dongha Lee, Jianxun Lian, Xing Xie, Hwanjo Yu

ACM The Web Conference (WWW), 2023

• We propose HetComp to compress ensemble of heterogeneous models, reducing huge inference costs while retaining high accuracy.

Learning Topology-Specific Experts for Molecular Property Prediction [link]

Suyeon Kim, Dongha Lee, SeongKu Kang, Seonghyeon Lee, Hwanjo Yu

AAAI Conference on Artificial Intelligence (AAAI), 2023

• We introduce a new topology-based gating module for molecular property prediction.

Consensus Learning from Heterogeneous Objectives for One-Class Collaborative Filtering [link]

SeongKu Kang, Dongha Lee, Wonbin Kweon, Junyoung Hwang, Hwanjo Yu

ACM The Web Conference (WWW), 2022

• We introduce a new training strategy that exploits the complementarity from heterogeneous objectives for one-class collaborative filtering.

TaxoCom: Topic Taxonomy Completion with Hierarchical Discovery of Novel Topic Clusters [link]

Dongha Lee, Jiaming Shen, SeongKu Kang, Susik Yoon, Jiawei Han, Hwanjo Yu

ACM The Web Conference (WWW), 2022

• We introduce TaxoCom which recursively expands the topic taxonomy by discovering novel sub-topic clusters of terms and documents.

Obtaining Calibrated Probabilities with Personalized Ranking Models [link]

Wonbin Kweon, SeongKu Kang, Hwanjo Yu

AAAI Conference on Artificial Intelligence (AAAI), Oral, 2022.

• We propose two calibration methods for ranking model and a new unbiased empirical risk minimization framework to guide the calibration methods.

Mitigating viewpoint sensitivity of self-supervised one-class classifiers [link]

Hyunjun Ju, Dongha Lee, SeongKu Kang, Hwanjo Yu

Information Sciences (SCI), 2022

• We propose GROC, a one-class classifier robust to geometrically-transformed inputs.

Personalized Knowledge Distillation for Recommender System [link]

SeongKu Kang, Dongha Lee, Wonbin Kweon, Hwanjo Yu

Knowledge-Based Systems (SCI), 2022

• We introduce a new distillation strategy, distilling the preference knowledge in a balanced way without relying on any hyperparameter.

Topology Distillation for Recommender System [link]

SeongKu Kang, Junyoung Hwang, Wonbin Kweon, Hwanjo Yu
ACM SIGKDD Conf. on Knowledge Discovery and Data Mining (KDD), 2021

• We introduce Topology Distillation, which guides the student by transferring the topological structure built upon the relations in the teacher space.

Bootstrapping User and Item Representations for One-Class Collaborative Filtering [link]

Dongha Lee, SeongKu Kang, Hyunjun Ju, Chanyoung Park, Hwanjo Yu
ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 2021

• We propose BUIR, a new training framework that does not require negative sampling.

Unsupervised Proxy Selection for Session-based Recommender Systems [link]

Junsu Cho, SeongKu Kang, Dongmin Hyun, Hwanjo Yu
ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 2021

• We propose ProxySR which imitates the missing information of general user interest by modeling proxies of sessions.

Learning Heterogeneous Temporal Patterns of User Preference for Timely Recommendation [link]

Junsu Cho, Dongmin Hyun, SeongKu Kang, Hwanjo Yu

ACM International World-Wide Web Conference (WWW), 2021

• We propose TimelyRec which exploits heterogeneous temporal patterns of user preference.

Bidirectional Distillation for Top-K Recommender System [link]

Wonbin Kweon, SeongKu Kang, Hwanjo Yu

ACM International World-Wide Web Conference (WWW), 2021

• We introduce BD framework whereby both the teacher and the student collaboratively improve with each other.

Item-side Ranking Regularized Distillation for Recommender System [link]

SeongKu Kang, Junyoung Hwang, Wonbin Kweon, Hwanjo Yu

Information Sciences (SCI), 2021

• We propose a new regularization method designed to maximize the effect of the ranking distillation.

DE-RRD: A Knowledge Distillation Framework for Recommender System [link]

SeongKu Kang, Junyoung Hwang, Wonbin Kweon, Hwanjo Yu

ACM International Conference on Information and Knowledge Management (CIKM), 2020

• We propose two methods: (1) DE for latent knowledge distillation, (2) RRD for ranking knowledge distillation.

Deep Rating Elicitation for New Users in Collaborative Filtering [link]

Wonbin Kweon, SeongKu Kang, Junyoung Hwang , Hwanjo Yu

ACM International World-Wide Web Conference (WWW), Short paper, 2020

• We introduce DRE, a new framework to choose the initial seed items for new users.

Multi-Modal Component Embedding for Fake News Detection [link]

SeongKu Kang, Junyoung Hwang , Hwanjo Yu

IEEE International Conf. Ubiquitous Information Management and Communication (IMCOM), 2020

• We explore the multi-modal feature combination for fake news detection.

Semi-Supervised Learning for Cross-Domain Recommendation to Cold-start Users [link]

SeongKu Kang, Junyoung Hwang, Dongha Lee, Hwanjo Yu

ACM International Conference on Information and Knowledge Management (CIKM), 2019

• We introduce a semi-supervised learning method that is effective when the overlapping users are exteremly limited.

• Ranked 12th among the most influential papers at CIKM 2019 (link) 

Densifying a Trust Network for Effective Collaborative Filtering [link]

SeongKu Kang, Jemin Wang, Yeon-Chang Lee, Sang-Wook Kim

Korean DataBase Conference (KDBC), 🏆 Best Paper Award, 2017.

• We densify social network to provide supplementary signals for recommendation.