Papers by Sun Kim
From Speculation Detection to Trustworthy Relational Tuples in Information Extraction (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies on speculation detection are defined at sentence level, but not all factual tuples extracted from a sentence are speculative. |
| Approach: | They propose to study speculations in OIE tuples and determine whether a tample is speculative. |
| Outcome: | The proposed model is based on the LSOIE dataset and provides labels for speculative tuples. |
SLM as Guardian: Pioneering AI Safety with Small Language Model (2024.emnlp-industry)
Copied to clipboard
Ohjoon Kwon, Donghyeon Jeon, Nayoung Choi, Gyu-Hwung Cho, Hwiyeol Jo, Changbong Kim, Hyunwoo Lee, Inho Kang, Sun Kim, Taiwoo Park
| Challenge: | Prior safety research on large language models focused on aligning them to safety requirements, but internalizing such safeguard features into larger models brought challenges of higher training cost and unintended degradation of helpfulness. |
| Approach: | They propose a multi-task learning mechanism that integrates harmful query detection and safeguard response into a single model. |
| Outcome: | The proposed approach outperforms the publicly available LLMs in harmful query detection and safeguard response generation. |
LLMs Behind the Scenes: Enabling Narrative Scene Illustration (2025.emnlp-main)
Copied to clipboard
| Challenge: | Generative AI has established the ability to readily transform content from one medium to another. |
| Approach: | They propose a pipeline that uses large language models to prompt text-to-image models to generate scenes for story text. |
| Outcome: | The proposed pipeline synthesizes illustrations for scenes in a story corpus using human annotation tasks. |
Personality Vector: Modulating Personality of Large Language Models by Model Merging (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to induce personality in large language models (LLMs) fail to capture the continuous nature of human traits. |
| Approach: | They propose a method for personality modulation in large language models by model merging by subtracting weights of pre-trained models from those of fine-tuned models. |
| Outcome: | The proposed method allows LLMs to exhibit desired personality traits without additional training. |
PromptSource: An Integrated Development Environment and Repository for Natural Language Prompts (2022.acl-demo)
Copied to clipboard
Stephen Bach, Victor Sanh, Zheng Xin Yong, Albert Webson, Colin Raffel, Nihal V. Nayak, Abheesht Sharma, Taewoon Kim, M Saiful Bari, Thibault Fevry, Zaid Alyafeai, Manan Dey, Andrea Santilli, Zhiqing Sun, Srulik Ben-david, Canwen Xu, Gunjan Chhablani, Han Wang, Jason Fries, Maged Al-shaibani, Shanya Sharma, Urmish Thakker, Khalid Almubarak, Xiangru Tang, Dragomir Radev, Mike Tian-jian Jiang, Alexander Rush
| Challenge: | PromptSource is a system for creating, sharing, and using natural language prompts . prompts are used to train and query language models in zero-shot learning settings . |
| Approach: | PromptSource is a system for creating, sharing, and using natural language prompts . et al.: using prompts to train and query language models is emerging area in NLP . they propose a templating language for defining data-linked prompts, a user interface that iterates on prompt development . |
| Outcome: | PromptSource is a system for creating, sharing, and using natural language prompts . it has a templating language for defining data-linked prompts and a community-driven set of guidelines . |
DocOIE: A Document-level Context-Aware Dataset for OpenIE (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing solutions focus on extracting tuples at sentence level, but sentences exist as part of a document rather than standalone. |
| Approach: | They propose to annotate 800 sentences from 80 documents to form a DocOIE dataset . they propose to use document-level context to improve OpenIE performance . |
| Outcome: | The proposed OpenIE model improves performance by incorporating documentlevel context into the dataset. |
Taxonomy and Analysis of Sensitive User Queries in Generative AI Search System (2025.findings-naacl)
Copied to clipboard
Hwiyeol Jo, Taiwoo Park, Hyunwoo Lee, Nayoung Choi, Changbong Kim, Ohjoon Kwon, Donghyeon Jeon, Eui Hyeon Lee, Kyoungho Shin, Lim Sun Suk, Kyungmi Kim, Lee Jihye, Sun Kim
| Challenge: | generative LLMs have been used by industries for various purposes, but limited resources and limited experience hinder their deployment and maintenance. |
| Approach: | They propose a taxonomy for sensitive search queries and outline approaches to generating generative LLMs. |
| Outcome: | The proposed model can be used to analyze sensitive queries from real users. |
MV-CLAM: Multi-View Molecular Interpretation with Cross-Modal Projection via Language Model (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing molecule-text models ignore complementary information in different molecular views and rely on single-view representations, limiting molecule structural understanding. |
| Approach: | They propose a framework that aligns multi-view molecular representations into a unified textual space using a multi-querying transformer. |
| Outcome: | The proposed framework ensures cross-view consistency while the proposed token-level contrastive loss preserves diverse molecular features across textual queries. |
X-RiSAWOZ: High-Quality End-to-End Multilingual Dialogue Datasets and Few-shot Agents (2023.findings-acl)
Copied to clipboard
Mehrad Moradshahi, Tianhao Shen, Kalika Bali, Monojit Choudhury, Gael de Chalendar, Anmol Goel, Sungkyun Kim, Prashant Kodali, Ponnurangam Kumaraguru, Nasredine Semmar, Sina Semnani, Jiwon Seo, Vivek Seshadri, Manish Shrivastava, Michael Sun, Aditya Yadavalli, Chaobin You, Deyi Xiong, Monica Lam
| Challenge: | X-RiSAWOZ dataset has more than 18,000 human-verified dialogue utterances for each language . Xiaoping and Xinhui are the main challenges for task-oriented dialogue research . |
| Approach: | They develop a toolkit to accelerate the post-editing of a new language dataset after translation . their dataset, code, and toolkit are released open-source . |
| Outcome: | The proposed toolkit accelerates the post-editing of a new language dataset after translation. |
STEAM: A Semantic-Level Knowledge Editing Framework for Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for locate-and-editing focus on token-level likelihood optimization without addressing semantic coherence. |
| Approach: | They propose a semantic-level knowledge editing framework that enhances integration of updated knowledge into the model's knowledge structure. |
| Outcome: | The proposed framework improves integration of updated knowledge into the model's knowledge structure and improves semantic coherence. |
Kiss up, Kick down: Exploring Behavioral Changes in Multi-modal Large Language Models with Assigned Visual Personas (2024.emnlp-main)
Copied to clipboard
Seungjong Sun, Eungu Lee, Seo Baek, Seunghyun Hwang, Wonbyung Lee, Dongyan Nan, Bernard Jansen, Jang Kim
| Challenge: | Large language models (LLMs) exhibit a high degree of alignment with human behavior based on their robust capabilities for natural language understanding and generation. |
| Approach: | They developed a dataset of 5K fictional avatar images for assignment as visual personas to large language models (LLMs) and analyzed their negotiation behaviors based on the visual traits depicted in these images. |
| Outcome: | The proposed model exhibited aggressive negotiation behaviors when the opponent’s image appeared less aggressive than their own, and less aggressive negotiation behavior when the opposing image appeared more aggressive. |
Agent-GWO: Collaborative Agents for Dynamic Prompt Optimization in Large Language Models (2026.findings-acl)
Copied to clipboard
Xudong Wang, Chaoning Zhang, Chenghao Li, Shuxu Chen, Qigan Sun, Jiaquan Zhang, Fachrina Dewi Puspitasari, Tae-Ho Kim, Jiwei Wei, Malu Zhang, Guoqing Wang, Yang Yang, Heng Tao Shen
| Challenge: | Existing automatic prompt optimization methods fail to optimize prompts and decoding hyperparameters within a unified framework to achieve stable global improvements. |
| Approach: | They propose a dynamic prompt optimization framework for complex reasoning that unifies prompt templates and decodes hyperparameters as inheritable agent configurations. |
| Outcome: | Experiments on multiple mathematical and hybrid reasoning benchmarks show that Agent-GWO improves accuracy and stability over existing prompt optimization methods. |
Clinical Note Owns its Hierarchy: Multi-Level Hypergraph Neural Networks for Patient-Level Representation Learning (2023.acl-long)
Copied to clipboard
| Challenge: | Clinical notes of patient EHRs contain valuable information from healthcare professionals, but have been underutilized due to their difficult-to-understand contents and complex hierarchies. |
| Approach: | They propose to use clinical notes to learn more balanced knowledge from EHRs by assembling useful neutral words with rare keywords via note and taxonomy level hyperedges. |
| Outcome: | The proposed method can retain clinical semantic information by (1) frequent neutral words and (2) hierarchies with imbalanced distribution. |
On Measures of Biases and Harms in NLP (2022.findings-aacl)
Copied to clipboard
Sunipa Dev, Emily Sheng, Jieyu Zhao, Aubrie Amstutz, Jiao Sun, Yu Hou, Mattie Sanseverino, Jiin Kim, Akihiro Nishi, Nanyun Peng, Kai-Wei Chang
| Challenge: | Recent studies show that natural language processing (NLP) technologies propagate societal biases about demographic groups associated with attributes such as gender, race, and nationality. |
| Approach: | They propose a framework for harms and questions to help practitioners understand biases . they propose measurable measures to detect and mitigate biased groups . |
| Outcome: | The proposed framework provides a framework for harms and questions for practitioners to answer to guide the development of bias measures. |
Introducing Semantics into Speech Encoders (2023.acl-long)
Copied to clipboard
Derek Xu, Shuyan Dong, Changhan Wang, Suyoun Kim, Zhaojiang Lin, Bing Liu, Akshat Shrivastava, Shang-Wen Li, Liang-Hsuan Tseng, Guan-Ting Lin, Alexei Baevski, Hung-yi Lee, Yizhou Sun, Wei Wang
| Challenge: | Existing self-supervised speech encoders contain primarily acoustic rather than semantic information. |
| Approach: | They propose a task-agnostic unsupervised way to incorporate semantic information from large language model (LLM) systems into self-supervised speech encoders without labeled audio transcriptions. |
| Outcome: | The proposed approach improves spoken language understanding (SLU) performance by over 5% on intent classification (IC), with modest gains in named entity resolution (NER) and slot filling (SF), and spoken question answering (SQA) score by over 22%. |
Bridging the Code Gap: A Joint Learning Framework across Medical Coding Systems (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods for automating medical coding focus on a single coding system . however, there are still challenges to overcome in coding. |
| Approach: | They propose a joint learning framework for Across Medical coding systems which jointly learns different coding system through multi-task learning. |
| Outcome: | The proposed framework improves the performance of the MIMIC-IV ICD-9 and MIMICIV I CD-10 datasets. |
PrismRAG: Boosting RAG Factuality with Distractor Resilience and Strategized Reasoning (2025.emnlp-industry)
Copied to clipboard
Mohammad Kachuee, Teja Gollapudi, Minseok Kim, Yin Huang, Kai Sun, Xiao Yang, Jiaqi Wang, Nirav Shah, Yue Liu, Aaron Colak, Anuj Kumar, Wen-tau Yih, Xin Luna Dong
| Challenge: | Existing methods to improve factuality of large language models (LLMs) rely on human-engineered instructions. |
| Approach: | They propose a retrieval-augmented generation framework that trains the model with distractor-aware QA pairs mixing gold evidence with subtle distractor passages and instills reasoning-centric habits that make the LLM plan, rationalize, and synthesize without extensive human engineered instructions. |
| Outcome: | The proposed framework outperforms state-of-the-art solutions across 12 open-book RAG QA benchmarks and is being deployed in production. |
Anaphor Assisted Document-Level Relation Extraction (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for document-level relation extraction are incomplete and lack anaphor for identifying relations between entities. |
| Approach: | They propose an Anaphor-Assisted (AA) framework for document-level relation extraction . they use a document or sentences as intermediate nodes to model cross-sentence entity interactions . |
| Outcome: | The proposed framework achieves state-of-the-art on the widely-used datasets. |
Jailbreaking Multimodal Large Language Models using Multi-Clip Video (2026.acl-long)
Copied to clipboard
| Challenge: | Existing studies show that video inputs can bypass safety alignment, yet it remains unclear which properties of video input induce this vulnerability. |
| Approach: | They propose a simple image-based defense that mitigates the vulnerability of MLLMs by analyzing video inputs. |
| Outcome: | The proposed defense leverages the relative robustness of the image modality. |
Syntactic Multi-view Learning for Open Information Extraction (2022.emnlp-main)
Copied to clipboard
| Challenge: | Open Information Extraction (OpenIE) aims to generate structured tuples from unstructured open-domain text. |
| Approach: | They propose to model constituency and dependency trees into word-level graphs and combine them with sentential semantic representations to extract relational tuples. |
| Outcome: | The proposed model integrates constituency and dependency trees into word-level graphs and enables neural OpenIE to learn from syntactic structures. |
PHEE: A Dataset for Pharmacovigilance Event Extraction from Text (2022.emnlp-main)
Copied to clipboard
Zhaoyue Sun, Jiazheng Li, Gabriele Pergola, Byron Wallace, Bino John, Nigel Greene, Joseph Kim, Yulan He
| Challenge: | Using NLP methods to discover and extract adverse drug events from unstructured textual data is difficult because it requires time-consuming manual curation. |
| Approach: | They propose to use a hierarchical event schema to extract annotated events from medical case reports and biomedical literature to analyze patient data. |
| Outcome: | The proposed dataset is the largest public dataset to date and contains over 5000 events from medical case reports and biomedical literature. |
Open Information Extraction via Chunks (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing OIE systems split a sentence into tokens and recognize token spans as tuple relations and arguments. |
| Approach: | They propose to split a sentence into tokens and recognize token spans as tuple relations and arguments. |
| Outcome: | The proposed model achieves state-of-the-art on multiple OIE datasets showing that SaC has better properties than sentence as token sequence. |