Hyunsik Yoo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Hyunsik Yoo has a total of 36 co-authors. Here are the top co-authors by number of shared publications:
- Hanghang Tong
University of Illinois Urbana-Champaign, Urbana, IL, USA
6 shared publications - Ruizhong Qiu
University of Illinois Urbana-Champaign, Urbana, IL, USA
5 shared publications - SeongKu Kang
Korea University, Seoul, Republic of Korea
5 shared publications - Fei Wang
Amazon, Sunnyvale, CA, USA
3 shared publications - Caizhi Xu
Amazon.com, Inc., Seattle, WA, USA
2 shared publications - Hwanjo Yu
Pohang University of Science and Technology, Pohang, Republic of Korea
2 shared publications
Latest Publications
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 2503-2513
Beyond Observed Auxiliary Relations: Environment-Conditioned Modeling for Multi-Behavior Recommendation
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 2503-2513
Open AccessBeyond Observed Auxiliary Relations: Environment-Conditioned Modeling for Multi-Behavior Recommendation
- Open Access
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 2244-2255
TRACER: Balancing Stability-Plasticity-Cognitivity Trilemma for LLM Enhanced Continual Recommendation
Continual recommendation aims to capture evolving user interests from streaming data but struggles with sparsity. LLM enhancers mitigate this with semantic knowledge, but naive integration creates a new conflict. We identify this as the Stability-Plasticity-Cognitivity (SPC) Trilemma, where generalized LLM semantic priors (Cognitivity) conflict with retaining personalized historical preferences (Stability) and adapting to individual interest shifts (Plasticity). To address this, we propose Trilemma-Responsive Adaptive Continual Enhancement for Recommendation (TRACER). TRACER synergistically combines three specialized modules, each targeting stability, plasticity, or cognitivity, while preventing any single lemma from dominating. This holistic design enables semantic knowledge to support history retention and adaptation to evolving interests without disrupting continual learning. Across five real-world datasets, TRACER effectively harmonizes the SPC trilemma and outperforms state-of-the-art baselines by up to 14.38%. Our code is available at https://github.com/woo-joo/TRACER_CIKM26.
- 6 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 2244-2255
Open AccessTRACER: Balancing Stability-Plasticity-Cognitivity Trilemma for LLM Enhanced Continual Recommendation
Continual recommendation aims to capture evolving user interests from streaming data but struggles with sparsity. LLM enhancers mitigate this with semantic knowledge, but naive integration creates a new conflict. We identify this as the Stability-Plasticity-Cognitivity (SPC) Trilemma, where generalized LLM semantic priors (Cognitivity) conflict with retaining personalized historical preferences (Stability) and adapting to individual interest shifts (Plasticity). To address this, we propose Trilemma-Responsive Adaptive Continual Enhancement for Recommendation (TRACER). TRACER synergistically combines three specialized modules, each targeting stability, plasticity, or cognitivity, while preventing any single lemma from dominating. This holistic design enables semantic knowledge to support history retention and adaptation to evolving interests without disrupting continual learning. Across five real-world datasets, TRACER effectively harmonizes the SPC trilemma and outperforms state-of-the-art baselines by up to 14.38%. Our code is available at https://github.com/woo-joo/TRACER_CIKM26.
- Open Access
- 24 September 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 4 (4) , 1-27
SCOUT: Coupling-Free Bounds for Trillion-Scale Top-k Retrieval in Sparse Tensor Factorization
Ranked retrieval over massive implicit result spaces is a core query-processing problem in data management. A natural instance arises when using tensor factorization (TF) models for retrieval over sparse tensors, where candidate tuples correspond to unobserved higher-order interactions over a multiway Cartesian product and are scored by a learned TF model. Real-world tensors in domains such as biomedical analysis, recommendation, and temporal interaction modeling are sparse and incomplete, and the candidate space can reach the trillion scale, making exhaustive scoring and sorting infeasible. While most prior work focuses on improving TF training or model accuracy, scalable inference-time top- k ranked retrieval over learned multiway scores has been relatively underexplored. Moreover, existing pairwise top-k search techniques, such as Maximum Inner Product Search (MIPS), do not directly extend to TF due to coupled multiway scores and the combinatorial candidate space. We propose Scout, a scalable and compatible framework for top- k ranked retrieval over massive implicit tensor spaces. Scout reparameterizes TF scores into nonnegative per-mode potentials and a normalized interaction term, yielding a coupling-free product-form bound that enables ranked access, admissible pruning, and reliable candidate ordering, without mode-coupled materialization as in MIPS-style reductions. Using this bound, Scout performs an N -dimensional best-first traversal with deferred scoring, supporting budgeted anytime retrieval and certified stopping when exact top- k is required. Scout further verifies candidates in GPU-friendly block-wise batches, enabling high-throughput execution under memory and latency budgets. Experiments show that Scout achieves over 10 3 x speedup over full enumeration for various TF methods and up to 11x over MIPS-based baselines, while maintaining comparable retrieval quality.
- 24 September 2026
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Management of Data
- Vol. 4 (4) , 1-27
Open AccessSCOUT: Coupling-Free Bounds for Trillion-Scale Top-k Retrieval in Sparse Tensor Factorization
Ranked retrieval over massive implicit result spaces is a core query-processing problem in data management. A natural instance arises when using tensor factorization (TF) models for retrieval over sparse tensors, where candidate tuples correspond to unobserved higher-order interactions over a multiway Cartesian product and are scored by a learned TF model. Real-world tensors in domains such as biomedical analysis, recommendation, and temporal interaction modeling are sparse and incomplete, and the candidate space can reach the trillion scale, making exhaustive scoring and sorting infeasible. While most prior work focuses on improving TF training or model accuracy, scalable inference-time top- k ranked retrieval over learned multiway scores has been relatively underexplored. Moreover, existing pairwise top-k search techniques, such as Maximum Inner Product Search (MIPS), do not directly extend to TF due to coupled multiway scores and the combinatorial candidate space. We propose Scout, a scalable and compatible framework for top- k ranked retrieval over massive implicit tensor spaces. Scout reparameterizes TF scores into nonnegative per-mode potentials and a normalized interaction term, yielding a coupling-free product-form bound that enables ranked access, admissible pruning, and reliable candidate ordering, without mode-coupled materialization as in MIPS-style reductions. Using this bound, Scout performs an N -dimensional best-first traversal with deferred scoring, supporting budgeted anytime retrieval and certified stopping when exact top- k is required. Scout further verifies candidates in GPU-friendly block-wise batches, enabling high-throughput execution under memory and latency budgets. Experiments show that Scout achieves over 10 3 x speedup over full enumeration for various TF methods and up to 11x over MIPS-based baselines, while maintaining comparable retrieval quality.
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 6469-6480
Mixture of Sequence: Theme-Aware Mixture-of-Experts for Long-Sequence Recommendation
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 6469-6480
Open AccessMixture of Sequence: Theme-Aware Mixture-of-Experts for Long-Sequence Recommendation
- Top Cited PapersOpen Access
- 21 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 313-323
Capturing User Interests from Data Streams for Continual Sequential Recommendation
Transformer-based sequential recommendation (SR) models excel at modeling long-range dependencies, but suffer from high computational costs and catastrophic forgetting during continuous updates. Although continual learning has been applied to recommendation, existing methods gradually forget long-term user preferences and remain underexplored in SR. In this paper, we introduce C ontinual S equential T ransformer for Rec ommendation (CSTRec), which effectively adapt to current interests by leveraging preserved historical knowledge. Its core is Continual Sequential Attention (CSA), a linear attention tailored for continual SR, which partially retain historical knowledge without direct access to prior data. CSA features: (1) Cauchy-Schwarz Normalization to stabilize learning over time under uneven user interaction frequencies, and (2) Collaborative Interest Enrichment via shared, learnable interest pools to mitigate forgetting. We also introduce a new technique for new user adaptation by transferring historical knowledge from existing users with similar interests. Extensive experiments show CSTRec's superior performance in both knowledge retention and acquisition. Our code is available at https://github.com/Gyu-Seok0/CSTRec_WSDM26.
- 21 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 313-323
Top Cited PapersOpen AccessCapturing User Interests from Data Streams for Continual Sequential Recommendation
Transformer-based sequential recommendation (SR) models excel at modeling long-range dependencies, but suffer from high computational costs and catastrophic forgetting during continuous updates. Although continual learning has been applied to recommendation, existing methods gradually forget long-term user preferences and remain underexplored in SR. In this paper, we introduce C ontinual S equential T ransformer for Rec ommendation (CSTRec), which effectively adapt to current interests by leveraging preserved historical knowledge. Its core is Continual Sequential Attention (CSA), a linear attention tailored for continual SR, which partially retain historical knowledge without direct access to prior data. CSA features: (1) Cauchy-Schwarz Normalization to stabilize learning over time under uneven user interaction frequencies, and (2) Collaborative Interest Enrichment via shared, learnable interest pools to mitigate forgetting. We also introduce a new technique for new user adaptation by transferring historical knowledge from existing users with similar interests. Extensive experiments show CSTRec's superior performance in both knowledge retention and acquisition. Our code is available at https://github.com/Gyu-Seok0/CSTRec_WSDM26.
- Open Access
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 6857-6860
Continual Recommender Systems
- 10 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 6857-6860
Open AccessContinual Recommender Systems
- Open Access
- 20 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 1833-1843
Generalizable Recommender System During Temporal Popularity Distribution Shifts
- 20 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 1833-1843
Open AccessGeneralizable Recommender System During Temporal Popularity Distribution Shifts
- Open Access
- 13 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 2092-2101
Embracing Plasticity: Balancing Stability and Plasticity in Continual Recommender Systems
- 13 July 2025
- Published by Association for Computing Machinery (ACM)
- p. 2092-2101
Open AccessEmbracing Plasticity: Balancing Stability and Plasticity in Continual Recommender Systems
- Open Access
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 3667-3678
Ensuring User-side Fairness in Dynamic Recommender Systems
User-side group fairness is crucial for modern recommender systems, alleviating performance disparities among user groups defined by sensitive attributes like gender, race, or age. In the everevolving landscape of user-item interactions, continual adaptation to newly collected data is crucial for recommender systems to stay aligned with the latest user preferences. However, we observe that such continual adaptation often worsen performance disparities. This necessitates a thorough investigation into user-side fairness in dynamic recommender systems. This problem is challenging due to distribution shifts, frequent model updates, and nondifferentiability of ranking metrics. To our knowledge, this paper presents the first principled study on ensuring user-side fairness in dynamic recommender systems. We start with theoretical analyses on fine-tuning v.s. retraining, showing that the best practice is incremental fine-tuning with restart. Guided by our theoretical analyses, we propose FAir Dynamic rEcommender (FADE), an end-to-end fine-tuning framework to dynamically ensure user-side fairness over time. To overcome the non-differentiability of recommendation metrics in the fairness loss, we further introduce Differentiable Hit (DH) as an improvement over the recent NeuralNDCG method, not only alleviating its gradient vanishing issue but also achieving higher efficiency. Besides that, we also address the instability issue of the fairness loss by leveraging the competing nature between the recommendation loss and the fairness loss. Through extensive experiments on real-world datasets, we demonstrate that FADE effectively and efficiently reduces performance disparities with little sacrifice in the overall recommendation performance.
- 13 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 3667-3678
Open AccessEnsuring User-side Fairness in Dynamic Recommender Systems
User-side group fairness is crucial for modern recommender systems, alleviating performance disparities among user groups defined by sensitive attributes like gender, race, or age. In the everevolving landscape of user-item interactions, continual adaptation to newly collected data is crucial for recommender systems to stay aligned with the latest user preferences. However, we observe that such continual adaptation often worsen performance disparities. This necessitates a thorough investigation into user-side fairness in dynamic recommender systems. This problem is challenging due to distribution shifts, frequent model updates, and nondifferentiability of ranking metrics. To our knowledge, this paper presents the first principled study on ensuring user-side fairness in dynamic recommender systems. We start with theoretical analyses on fine-tuning v.s. retraining, showing that the best practice is incremental fine-tuning with restart. Guided by our theoretical analyses, we propose FAir Dynamic rEcommender (FADE), an end-to-end fine-tuning framework to dynamically ensure user-side fairness over time. To overcome the non-differentiability of recommendation metrics in the fairness loss, we further introduce Differentiable Hit (DH) as an improvement over the recent NeuralNDCG method, not only alleviating its gradient vanishing issue but also achieving higher efficiency. Besides that, we also address the instability issue of the fairness loss by leveraging the competing nature between the recommendation loss and the fairness loss. Through extensive experiments on real-world datasets, we demonstrate that FADE effectively and efficiently reduces performance disparities with little sacrifice in the overall recommendation performance.
- Open Access
- 30 April 2023
- Published by Association for Computing Machinery (ACM)
- p. 231-239
Disentangling Degree-related Biases and Interest for Out-of-Distribution Generalized Directed Network Embedding
The goal of directed network embedding is to represent the nodes in a given directed network as embeddings that preserve the asymmetric relationships between nodes. While a number of directed network embedding methods have been proposed, we empirically show that the existing methods lack out-of-distribution generalization abilities against degree-related distributional shifts. To mitigate this problem, we propose ODIN (Out-of-Distribution Generalized Directed Network Embedding), a new directed NE method where we model multiple factors in the formation of directed edges. Then, for each node, ODIN learns multiple embeddings, each of which preserves its corresponding factor, by disentangling interest factors and biases related to in- and out-degrees of nodes. Our experiments on four real-world directed networks demonstrate that disentangling multiple factors enables ODIN to yield out-of-distribution generalized embeddings that are consistently effective under various degrees of shifts in degree distributions. Specifically, ODIN universally outperforms 9 state-of-the-art competitors in 2 LP tasks on 4 real-world datasets under both identical distribution (ID) and non-ID settings. The code is available at https://github.com/hsyoo32/odin.
- 30 April 2023
- Published by Association for Computing Machinery (ACM)
- p. 231-239
Open AccessDisentangling Degree-related Biases and Interest for Out-of-Distribution Generalized Directed Network Embedding
The goal of directed network embedding is to represent the nodes in a given directed network as embeddings that preserve the asymmetric relationships between nodes. While a number of directed network embedding methods have been proposed, we empirically show that the existing methods lack out-of-distribution generalization abilities against degree-related distributional shifts. To mitigate this problem, we propose ODIN (Out-of-Distribution Generalized Directed Network Embedding), a new directed NE method where we model multiple factors in the formation of directed edges. Then, for each node, ODIN learns multiple embeddings, each of which preserves its corresponding factor, by disentangling interest factors and biases related to in- and out-degrees of nodes. Our experiments on four real-world directed networks demonstrate that disentangling multiple factors enables ODIN to yield out-of-distribution generalized embeddings that are consistently effective under various degrees of shifts in degree distributions. Specifically, ODIN universally outperforms 9 state-of-the-art competitors in 2 LP tasks on 4 real-world datasets under both identical distribution (ID) and non-ID settings. The code is available at https://github.com/hsyoo32/odin.