Sungho Park

Data Systems Lab

POSTECH, South Korea

shpark@dblab.postech.ac.kr

I am a Ph.D. student at Data Systems Lab @ Pohang University of Science and Technology (POSTECH), under the supervision of Professor Wook-Shin Han. My research interests range from neural information retrieval to multi-modal open-domain question answering (such as open question answering over tables and text), and to Agentic AI.

Recently, I have focused on designing algorithms that retrieve and integrate tables and passages relevant to a query by leveraging semantic relationships across modalities. My work has been published in top-tier venues including ACL 2025 Main and EMNLP 2025 Main.

Research Focus

Neural Information Retrieval: Developing advanced retrieval systems that understand semantic relationships
Multi-modal Question Answering: Bridging structured (tables) and unstructured (text) data for comprehensive QA
Agentic AI: Enhancing language models with dynamic knowledge retrieval

Recent Achievements

ACL 2025 Main: HELIOS paper on multi-granular table-text retrieval
EMNLP 2025 Main: SAFE paper on schema-driven knowledge graph querying
KDD Cup 2024: Won multiple prizes in Meta CRAG Challenge for retrieval-augmented generation

news

May 16, 2025	🎉 My first submission to ACL is accepted! HELIOS: Harmonizing Early Fusion, Late Fusion, and LLM Reasoning for Multi-Granular Table-Text Retrieval achieves SOTA performance in RAG.
Jul 24, 2024	🏆 Won multiple prizes in KDD Cup 2024 Meta CRAG Challenge for retrieval-augmented generation! Our team “dRAGonRAnGers” achieved first place in comparison questions and post-processing tasks.
Apr 25, 2023	🎓 Got an internship offer from Oracle Labs (Switzerland) for summer 2023! Will be working on large language models. Congratulations to myself! 🚀

selected publications

SPARTA: Scalable and Principled Benchmark of Tree-Structured Multi-hop QA over Text and Tables

Sungho Park, Juen Kim, and Wook-Shin Han

Submitted, 2025

Abs Bib

SPARTA is a fully automated SQL-centric pipeline that constructs a tree-structured multi-hop QA benchmark by unifying structured and unstructured evidence from tables and text into a single relational representation.
@article{park2024sparta, title = {SPARTA: Scalable and Principled Benchmark of Tree-Structured Multi-hop QA over Text and Tables}, author = {Park, Sungho and Kim, Juen and Han, Wook-Shin}, year = {2025}, journal = {Submitted}, }
EMNLP
SAFE: Schema-Driven Approximate Distance Join for Efficient Knowledge Graph Querying

Sangoh Lee, Sungho Park, and Wook-Shin Han

2025

Abs Bib

SAFE introduces a schema-driven approximate distance join algorithm that refines noisy LLM-generated query graphs using schema-level constraints and efficiently aligns them with large knowledge graphs, enabling robust and scalable knowledge graph querying.
@article{lee2025safe, title = {SAFE: Schema-Driven Approximate Distance Join for Efficient Knowledge Graph Querying}, author = {Lee, Sangoh and Park, Sungho and Han, Wook-Shin}, booktitle = {Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing}, year = {2025}, publisher = {Association for Computational Linguistics}, }
ACL
HELIOS: Harmonizing Early Fusion, Late Fusion, and LLM Reasoning for Multi-Granular Table-Text Retrieval

Sungho Park, Joohyung Yun, Jongwuk Lee, and 1 more author

In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025

Abs Bib

HELIOS formulates retrieval as finding a query-relevant subgraph within a bipartite data graph built via early fusion of table segments and passages, and introduces a three-stage pipeline integrating early fusion, late fusion, and LLM reasoning.
@inproceedings{park2025helios, title = {HELIOS: Harmonizing Early Fusion, Late Fusion, and LLM Reasoning for Multi-Granular Table-Text Retrieval}, author = {Park, Sungho and Yun, Joohyung and Lee, Jongwuk and Han, Wook-Shin}, booktitle = {Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)}, pages = {32424--32444}, year = {2025}, }