Jongwuk Lee

Sunkyung Lee, Minjin Choi, Jongwuk Lee

Generative retrieval shed light on a new paradigm of document retrieval, aiming to directly generate the identifier of a relevant document for a query. While it takes advantage of bypassing the construction of auxiliary index structures, existing studies face two significant challenges: (i) the discrepancy between the knowledge of pre-trained language models and identifiers and (ii) the gap between training and inference that poses difficulty in learning to rank. To overcome these challenges, we propose a novel generative retrieval method, namely Generative retrieval via LExical iNdex learning (GLEN). For training, GLEN effectively exploits a dynamic lexical identifier using a two-phase index learning strategy, enabling it to learn meaningful lexical identifiers and relevance signals between queries and documents. For inference, GLEN utilizes collision-free inference, using identifier weights to rank documents without additional overhead. Experimental results prove that GLEN achieves state-of-the-art or competitive performance against existing generative retrieval methods on various benchmark datasets, e.g., NQ320k, MS MARCO, and BEIR. The code is available at https://github.com/skleee/GLEN.

Yoonjin Im, Eunseong Choi, Heejin Kook et al.

Knowledge Tracing (KT) aims to track proficiency based on a question-solving history, allowing us to offer a streamlined curriculum. Recent studies actively utilize attention-based mechanisms to capture the correlation between questions and combine it with the learner's characteristics for responses. However, our empirical study shows that existing attention-based KT models neglect the learner's forgetting behavior, especially as the interaction history becomes longer. This problem arises from the bias that overprioritizes the correlation of questions while inadvertently ignoring the impact of forgetting behavior. This paper proposes a simple-yet-effective solution, namely Forgetting-aware Linear Bias (FoLiBi), to reflect forgetting behavior as a linear bias. Despite its simplicity, FoLiBi is readily equipped with existing attentive KT models by effectively decomposing question correlations with forgetting behavior. FoLiBi plugged with several KT models yields a consistent improvement of up to 2.58% in AUC over state-of-the-art KT models on four benchmark datasets.

Dongcheol Lee, Hye-young Kim, Jongwuk Lee

Recent advances in the LLM-as-Extractor paradigm leverage large language models (LLMs) to transfer semantically rich item embeddings into sequential recommendation (SR) backbones. However, LLM-generated embeddings often suffer from strong anisotropy. Most vectors are concentrated in similar directions, resulting in a geometric imbalance that makes it difficult to adapt to collaborative signals during fine-tuning. To address this challenge, we propose Anisotropy-Controllable Embedding (ACE), which explicitly controls the anisotropy of LLM-generated embeddings. Specifically, ACE utilizes a linear autoencoder (LAE) to reshape the embedding distribution while preserving its semantic structure. In this process, the L2-regularization term mitigates the anisotropy by controlling the dispersion of embedding dimensions, while the reconstruction loss maintains semantic relationships among items. That is, ACE balances geometric uniformity and semantic embedding preservation for more stable learning. Extensive experiments demonstrate that ACE consistently outperforms existing LLM-enhanced SR models, yielding improvements of up to 12.4% and 11.8% in Recall@20 and NDCG@20, respectively.

Jae-woong Lee, Seongmin Park, Joonseok Lee et al.

Implicit feedback has been widely used to build commercial recommender systems. Because observed feedback represents users' click logs, there is a semantic gap between true relevance and observed feedback. More importantly, observed feedback is usually biased towards popular items, thereby overestimating the actual relevance of popular items. Although existing studies have developed unbiased learning methods using inverse propensity weighting (IPW) or causal reasoning, they solely focus on eliminating the popularity bias of items. In this paper, we propose a novel unbiased recommender learning model, namely BIlateral SElf-unbiased Recommender (BISER), to eliminate the exposure bias of items caused by recommender models. Specifically, BISER consists of two key components: (i) self-inverse propensity weighting (SIPW) to gradually mitigate the bias of items without incurring high computational costs; and (ii) bilateral unbiased learning (BU) to bridge the gap between two complementary models in model predictions, i.e., user- and item-based autoencoders, alleviating the high variance of SIPW. Extensive experiments show that BISER consistently outperforms state-of-the-art unbiased recommender models over several datasets, including Coat, Yahoo! R3, MovieLens, and CiteULike.

Seongmin Park, Mincheol Yoon, Jae-woong Lee et al.

Collaborative filtering (CF) is a pivotal technique in modern recommender systems. The learning process of CF models typically consists of three components: interaction encoder, loss function, and negative sampling. Although many existing studies have proposed various CF models to design sophisticated interaction encoders, recent work shows that simply reformulating the loss functions can achieve significant performance gains. This paper delves into analyzing the relationship among existing loss functions. Our mathematical analysis reveals that the previous loss functions can be interpreted as alignment and uniformity functions: (i) the alignment matches user and item representations, and (ii) the uniformity disperses user and item distributions. Inspired by this analysis, we propose a novel loss function that improves the design of alignment and uniformity considering the unique patterns of datasets called Margin-aware Alignment and Weighted Uniformity (MAWU). The key novelty of MAWU is two-fold: (i) margin-aware alignment (MA) mitigates user/item-specific popularity biases, and (ii) weighted uniformity (WU) adjusts the significance between user and item uniformities to reflect the inherent characteristics of datasets. Extensive experimental results show that MF and LightGCN equipped with MAWU are comparable or superior to state-of-the-art CF models with various loss functions on three public datasets.

Hyunsoo Kim, Jaewan Moon, Seongmin Park et al.

Modern recommender systems trained on domain-specific data often struggle to generalize across multiple domains. Cross-domain sequential recommendation has emerged as a promising research direction to address this challenge; however, existing approaches face fundamental limitations, such as reliance on overlapping users or items across domains, or unrealistic assumptions that ignore privacy constraints. In this work, we propose a new framework, MergeRec, based on model merging under a new and realistic problem setting termed data-isolated cross-domain sequential recommendation, where raw user interaction data cannot be shared across domains. MergeRec consists of three key components: (1) merging initialization, (2) pseudo-user data construction, and (3) collaborative merging optimization. First, we initialize a merged model using training-free merging techniques. Next, we construct pseudo-user data by treating each item as a virtual sequence in each domain, enabling the synthesis of meaningful training samples without relying on real user interactions. Finally, we optimize domain-specific merging weights through a joint objective that combines a recommendation loss, which encourages the merged model to identify relevant items, and a distillation loss, which transfers collaborative filtering signals from the fine-tuned source models. Extensive experiments demonstrate that MergeRec not only preserves the strengths of the original models but also significantly enhances generalizability to unseen domains. Compared to conventional model merging methods, MergeRec consistently achieves superior performance, with average improvements of up to 17.21% in Recall@10, highlighting the potential of model merging as a scalable and effective approach for building universal recommender systems. The source code is available at https://github.com/DIALLab-SKKU/MergeRec.

Seongmin Park, Mincheol Yoon, Minjin Choi et al.

In sequential recommendation (SR), neural models have been actively explored due to their remarkable performance, but they suffer from inefficiency inherent to their complexity. On the other hand, linear SR models exhibit high efficiency and achieve competitive or superior accuracy compared to neural models. However, they solely deal with the sequential order of items (i.e., sequential information) and overlook the actual timestamp (i.e., temporal information). It is limited to effectively capturing various user preference drifts over time. To address this issue, we propose a novel linear SR model, named TemporAl LinEar item-item model (TALE), incorporating temporal information while preserving training/inference efficiency, with three key components. (i) Single-target augmentation concentrates on a single target item, enabling us to learn the temporal correlation for the target item. (ii) Time interval-aware weighting utilizes the actual timestamp to discern the item correlation depending on time intervals. (iii) Trend-aware normalization reflects the dynamic shift of item popularity over time. Our empirical studies show that TALE outperforms ten competing SR models by up to 18.71% gains on five benchmark datasets. It also exhibits remarkable effectiveness in evaluating long-tail items by up to 30.45% gains. The source code is available at https://github.com/psm1206/TALE.

Sunkyung Lee, Jihye Back, Donghyeon Jeon et al.

Generative information retrieval (GenIR) formulates the retrieval process as a text-to-text generation task, leveraging the vast knowledge of large language models. However, existing works primarily optimize for relevance while often overlooking document trustworthiness. This is critical in high-stakes domains like healthcare and finance, where relying solely on semantic relevance risks retrieving unreliable information. To address this, we propose an Authority-aware Generative Retriever (AuthGR), the first framework that incorporates authority into GenIR. AuthGR consists of three key components: (i) Multimodal Authority Scoring, which employs a vision-language model to quantify authority from textual and visual cues; (ii) a Three-stage Training Pipeline to progressively instill authority awareness into the retriever; and (iii) a Hybrid Ensemble Pipeline for robust deployment. Offline evaluations demonstrate that AuthGR successfully enhances both authority and accuracy, with our 3B model matching a 14B baseline. Crucially, large-scale online A/B tests and human evaluations conducted on the commercial web search platform confirm significant improvements in real-world user engagement and reliability.

Sunkyung Lee, Minjin Choi, Eunseong Choi et al.

Generative recommendation is an emerging paradigm that leverages the extensive knowledge of large language models by formulating recommendations into a text-to-text generation task. However, existing studies face two key limitations in (i) incorporating implicit item relationships and (ii) utilizing rich yet lengthy item information. To address these challenges, we propose a Generative Recommender via semantic-Aware Multi-granular late fusion (GRAM), introducing two synergistic innovations. First, we design semantic-to-lexical translation to encode implicit hierarchical and collaborative item relationships into the vocabulary space of LLMs. Second, we present multi-granular late fusion to integrate rich semantics efficiently with minimal information loss. It employs separate encoders for multi-granular prompts, delaying the fusion until the decoding stage. Experiments on four benchmark datasets show that GRAM outperforms eight state-of-the-art generative recommendation models, achieving significant improvements of 11.5-16.0% in Recall@5 and 5.3-13.6% in NDCG@5. The source code is available at https://github.com/skleee/GRAM.

Sunkyung Lee, Seongmin Park, Jonghyo Kim et al.

Generative recommendation has emerged as a promising paradigm that formulates the recommendations into a text-to-text generation task, harnessing the vast knowledge of large language models. However, existing studies focus on considering the sequential order of items and neglect to handle the temporal dynamics across items, which can imply evolving user preferences. To address this limitation, we propose a novel model, Generative Recommender Using Time awareness (GRUT), effectively capturing hidden user preferences via various temporal signals. We first introduce Time-aware Prompting, consisting of two key contexts. The user-level temporal context models personalized temporal patterns across timestamps and time intervals, while the item-level transition context provides transition patterns across users. We also devise Trend-aware Inference, a training-free method that enhances rankings by incorporating trend information about items with generation likelihood. Extensive experiments demonstrate that GRUT outperforms state-of-the-art models, with gains of up to 15.4% and 14.3% in Recall@5 and NDCG@5 across four benchmark datasets. The source code is available at https://github.com/skleee/GRUT.

Jaewan Moon, Seongmin Park, Jongwuk Lee

Large language models (LLMs) have been widely adopted to enrich the semantic representation of textual item information in recommender systems. However, existing linear autoencoders (LAEs) that incorporate textual information rely on sparse word co-occurrence patterns, limiting their ability to capture rich textual semantics. To address this, we propose L3AE, the first integration of LLMs into the LAE framework. L3AE effectively integrates the heterogeneous knowledge of textual semantics and user-item interactions through a two-phase optimization strategy. (i) L3AE first constructs a semantic item-to-item correlation matrix from LLM-derived item representations. (ii) It then learns an item-to-item weight matrix from collaborative signals while distilling semantic item correlations as regularization. Notably, each phase of L3AE is optimized through closed-form solutions, ensuring global optimality and computational efficiency. Extensive experiments demonstrate that L3AE consistently outperforms state-of-the-art LLM-enhanced models on three benchmark datasets, achieving gains of 27.6% in Recall@20 and 39.3% in NDCG@20. The source code is available at https://github.com/jaewan7599/L3AE_CIKM2025.

Eunseong Choi, Hyeri Lee, Jongwuk Lee

In Open-domain Question Answering (ODQA), it is essential to discern relevant contexts as evidence and avoid spurious ones among retrieved results. The model architecture that uses concatenated multiple contexts in the decoding phase, i.e., Fusion-in-Decoder, demonstrates promising performance but generates incorrect outputs from seemingly plausible contexts. To address this problem, we propose the Multi-Granularity guided Fusion-in-Decoder (MGFiD), discerning evidence across multiple levels of granularity. Based on multi-task learning, MGFiD harmonizes passage re-ranking with sentence classification. It aggregates evident sentences into an anchor vector that instructs the decoder. Additionally, it improves decoding efficiency by reusing the results of passage re-ranking for passage pruning. Through our experiments, MGFiD outperforms existing models on the Natural Questions (NQ) and TriviaQA (TQA) datasets, highlighting the benefits of its multi-granularity solution.

Eunseong Choi, June Park, Hyeri Lee et al.

Retrieval-augmented generation (RAG) enhances the capabilities of large language models (LLMs) by incorporating external knowledge into their input prompts. However, when the retrieved context contradicts the LLM's parametric knowledge, it often fails to resolve the conflict between incorrect external context and correct parametric knowledge, known as context-memory conflict. To tackle this problem, we introduce Conflict-Aware REtrieval-Augmented Generation (CARE), consisting of a context assessor and a base LLM. The context assessor encodes compact memory token embeddings from raw context tokens. Through grounded/adversarial soft prompting, the context assessor is trained to discern unreliable context and capture a guidance signal that directs reasoning toward the more reliable knowledge source. Extensive experiments show that CARE effectively mitigates context-memory conflicts, leading to an average performance gain of 5.0\% on QA and fact-checking benchmarks, establishing a promising direction for trustworthy and adaptive RAG systems.

Jaewan Moon, Hye-young Kim, Jongwuk Lee

Linear autoencoder models learn an item-to-item weight matrix via convex optimization with L2 regularization and zero-diagonal constraints. Despite their simplicity, they have shown remarkable performance compared to sophisticated non-linear models. This paper aims to theoretically understand the properties of two terms in linear autoencoders. Through the lens of singular value decomposition (SVD) and principal component analysis (PCA), it is revealed that L2 regularization enhances the impact of high-ranked PCs. Meanwhile, zero-diagonal constraints reduce the impact of low-ranked PCs, leading to performance degradation for unpopular items. Inspired by this analysis, we propose simple-yet-effective linear autoencoder models using diagonal inequality constraints, called Relaxed Linear AutoEncoder (RLAE) and Relaxed Denoising Linear AutoEncoder (RDLAE). We prove that they generalize linear autoencoders by adjusting the degree of diagonal constraints. Experimental results demonstrate that our models are comparable or superior to state-of-the-art linear and non-linear models on six benchmark datasets; they significantly improve the accuracy of long-tail items. These results also support our theoretical insights on regularization and diagonal constraints in linear autoencoders.

Jae-woong Lee, Seongmin Park, Mincheol Yoon et al.

Because implicit user feedback for the collaborative filtering (CF) models is biased toward popular items, CF models tend to yield recommendation lists with popularity bias. Previous studies have utilized inverse propensity weighting (IPW) or causal inference to mitigate this problem. However, they solely employ pointwise or pairwise loss functions and neglect to adopt a contrastive loss function for learning meaningful user and item representations. In this paper, we propose Unbiased ConTrastive Representation Learning (uCTRL), optimizing alignment and uniformity functions derived from the InfoNCE loss function for CF models. Specifically, we formulate an unbiased alignment function used in uCTRL. We also devise a novel IPW estimation method that removes the bias of both users and items. Despite its simplicity, uCTRL equipped with existing CF models consistently outperforms state-of-the-art unbiased recommender models, up to 12.22% for Recall@20 and 16.33% for NDCG@20 gains, on four benchmark datasets.

Minjin Choi, Jinhong Kim, Joonsek Lee et al.

Session-based recommendation (SR) predicts the next items from a sequence of previous items consumed by an anonymous user. Most existing SR models focus only on modeling intra-session characteristics but pay less attention to inter-session relationships of items, which has the potential to improve accuracy. Another critical aspect of recommender systems is computational efficiency and scalability, considering practical feasibility in commercial applications. To account for both accuracy and scalability, we propose a novel session-based recommendation with a random walk, namely S-Walk. Precisely, S-Walk effectively captures intra- and inter-session correlations by handling high-order relationships among items using random walks with restart (RWR). By adopting linear models with closed-form solutions for transition and teleportation matrices that constitute RWR, S-Walk is highly efficient and scalable. Extensive experiments demonstrate that S-Walk achieves comparable or state-of-the-art performance in various metrics on four benchmark datasets. Moreover, the model learned by S-Walk can be highly compressed without sacrificing accuracy, conducting two or more orders of magnitude faster inference than existing DNN-based models, making it suitable for large-scale commercial systems.

Seungho Lee, Minhyun Lee, Jongwuk Lee et al.

Existing studies in weakly-supervised semantic segmentation (WSSS) using image-level weak supervision have several limitations: sparse object coverage, inaccurate object boundaries, and co-occurring pixels from non-target objects. To overcome these challenges, we propose a novel framework, namely Explicit Pseudo-pixel Supervision (EPS), which learns from pixel-level feedback by combining two weak supervisions; the image-level label provides the object identity via the localization map and the saliency map from the off-the-shelf saliency detection model offers rich boundaries. We devise a joint training strategy to fully utilize the complementary relationship between both information. Our method can obtain accurate object boundaries and discard co-occurring pixels, thereby significantly improving the quality of pseudo-masks. Experimental results show that the proposed method remarkably outperforms existing methods by resolving key challenges of WSSS and achieves the new state-of-the-art performance on both PASCAL VOC 2012 and MS COCO 2014 datasets.

Minjin Choi, Sunkyung Lee, Eunseong Choi et al.

Automated metaphor detection is a challenging task to identify metaphorical expressions of words in a sentence. To tackle this problem, we adopt pre-trained contextualized models, e.g., BERT and RoBERTa. To this end, we propose a novel metaphor detection model, namely metaphor-aware late interaction over BERT (MelBERT). Our model not only leverages contextualized word representation but also benefits from linguistic metaphor identification theories to distinguish between the contextual and literal meaning of words. Our empirical results demonstrate that MelBERT outperforms several strong baselines on four benchmark datasets, i.e., VUA-18, VUA-20, MOH-X, and TroFi.

Minjin Choi, jinhong Kim, Joonseok Lee et al.

Session-based recommendation aims at predicting the next item given a sequence of previous items consumed in the session, e.g., on e-commerce or multimedia streaming services. Specifically, session data exhibits some unique characteristics, i.e., session consistency and sequential dependency over items within the session, repeated item consumption, and session timeliness. In this paper, we propose simple-yet-effective linear models for considering the holistic aspects of the sessions. The comprehensive nature of our models helps improve the quality of session-based recommendation. More importantly, it provides a generalized framework for reflecting different perspectives of session data. Furthermore, since our models can be solved by closed-form solutions, they are highly scalable. Experimental results demonstrate that the proposed linear models show competitive or state-of-the-art performance in various metrics on several real-world datasets.

Minjin Choi, Yoonki Jeong, Joonseok Lee et al.

Top-N recommendation is a challenging problem because complex and sparse user-item interactions should be adequately addressed to achieve high-quality recommendation results. The local latent factor approach has been successfully used with multiple local models to capture diverse user preferences with different sub-communities. However, previous studies have not fully explored the potential of local models, and failed to identify many small and coherent sub-communities. In this paper, we present Local Collaborative Autoencoders (LOCA), a generalized local latent factor framework. Specifically, LOCA adopts different neighborhood ranges at the training and inference stages. Besides, LOCA uses a novel sub-community discovery method, maximizing the coverage of a union of local models and employing a large number of diverse local models. By adopting autoencoders as the base model, LOCA captures latent non-linear patterns representing meaningful user-item interactions within sub-communities. Our experimental results demonstrate that LOCA is scalable and outperforms state-of-the-art models on several public benchmarks, by 2.99~4.70% in Recall and 1.02~7.95% in NDCG, respectively.

Jae-woong Lee, Minjin Choi, Jongwuk Lee et al.

Knowledge distillation (KD) is a well-known method to reduce inference latency by compressing a cumbersome teacher model to a small student model. Despite the success of KD in the classification task, applying KD to recommender models is challenging due to the sparsity of positive feedback, the ambiguity of missing feedback, and the ranking problem associated with the top-N recommendation. To address the issues, we propose a new KD model for the collaborative filtering approach, namely collaborative distillation (CD). Specifically, (1) we reformulate a loss function to deal with the ambiguity of missing feedback. (2) We exploit probabilistic rank-aware sampling for the top-N recommendation. (3) To train the proposed model effectively, we develop two training strategies for the student model, called the teacher- and the student-guided training methods, selecting the most useful feedback from the teacher model. Via experimental results, we demonstrate that the proposed model outperforms the state-of-the-art method by 2.7-33.2% and 2.7-29.1% in hit rate (HR) and normalized discounted cumulative gain (NDCG), respectively. Moreover, the proposed model achieves the performance comparable to the teacher model.