Preprints

Leveraging Historical and Current Interests for Continual Sequential Recommendation [link]

Gyuseok Lee, Hyunsik Yoo, Junyoung Hwang, SeongKu Kang† , Hwanjo Yu†

• We propose CSTRec which continuously updates a transformer-based SR model with non-stationary data streams. 

LLM-Based Compact Reranking with Document Features for Scientific Retrieval [link]

Runchu Tian, Xueqiang Xu, Bowen Jin, SeongKu Kang, Jiawei Han

• We propose CORANK, a training-free, model-agnostic reranking framework for scientific retrieval.

DeepRetrieval: Hacking Real Search Engines and Retrievers with Large Language Models via Reinforcement Learning [link]

Pengcheng Jiang, Jiacheng Lin, Lang Cao, Runchu Tian, SeongKu Kang, Zifeng Wang, Jimeng Sun, Jiawei Han

• We propose DeepRetrieval, an RL-based framework for training LLMs to enhance information retrieval through query generation/rewriting.

Scientific Paper Retrieval with LLM-Guided Semantic-Based Ranking [link]

Yunyi Zhang, Ruozhen Yang, Siqi Jiao, SeongKu Kang, Jiawei Han

• We propose SemRank, LLM-Guided Semantic-Based Ranking, a plug-and-play framework for scientific paper retrieval.

Imagine All The Relevance: Scenario-Profiled Indexing with Knowledge Expansion for Dense Retrieval [link]

Sangam Lee, Ryang Heo, SeongKu Kang, Dongha Lee

• We propose SPIKE, a dense retrieval framework that explicitly indexes implicit relevance by decomposing documents into scenario-based retrieval unit.

Graph Signal Processing for Cross-Domain Recommendation [link]

Jeongeun Lee, SeongKu Kang, Won-Yong Shin, Jeongwhan Choi, Noseong Park, Dongha Lee

• We propose CGSP, a unified cross-domain recommendation framework based on graph signal processing.

Why These Documents? Explainable Generative Retrieval with Hierarchical Category Paths [link]

Sangam Lee, Ryang Heo, SeongKu Kang, Susik Yoon, Jinyoung Yeo, Dongha Lee

• We propose HyPE, which leverages hierarchical category paths as explanation, progressing from broad to specific semantic categories. 

SC-Rec: Enhancing Generative Retrieval with Self-Consistent Reranking for Sequential Recommendation [link]

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

• We propose SCREC, a unified recommender system that learns diverse preference knowledge from two distinct item indices and multiple prompt templates.