Grace Byun
Portrait of Grace Byun
GB

Grace (Sungjoo) Byun

PhD Candidate, Computer Science & Informatics — Emory University

My research asks how language models fail when it matters most — then builds the benchmarks, safety evaluations, and retrieval systems to catch it first.

GitHub LinkedIn Google Scholar

About

I'm a PhD candidate in Computer Science and Informatics at Emory University, working with Prof. Jinho D. Choi in the EmoryNLP Lab. My research sits at the intersection of language and safety: constructing clinician-annotated benchmarks for crisis detection, studying failure modes like sycophancy in multi-turn dialogue, and building retrieval-augmented systems that hold up against adversarial and proprietary document collections. I interned on the Search and Embeddings team at Cohere, working on context-pruning rerankers.

Before Emory, I completed a master's in Computational Linguistics at Seoul National University, where my dissertation focused on simplifying LLM alignment and detoxification through instruction and preference data, and a BA in Linguistics at Korea University.

Education

Sep 2024 – present

Emory University

PhD, Computer Science and Informatics

Advised by Prof. Jinho D. Choi, EmoryNLP Lab.

Sep 2022 – Aug 2024

Seoul National University

Master's, Computational Linguistics · GPA 4.2 / 4.3

Dissertation: Simplifying Large Language Model Alignment and Detoxification — Comprehensive Instruction and Preference Data Solutions.

Mar 2018 – Aug 2022

Korea University

BA, Linguistics · GPA 4.35 / 4.5 (Linguistics 4.5 / 4.5)

Experience

Jan 2026 – Aug 2026

Cohere

Search & Embeddings Team · Intern
  • Developed evaluation datasets and metrics for context-pruning rerankers to measure retrieval efficiency and relevance.
  • Designed and built a context-pruning reranker model to improve retrieval precision and reduce redundant context.
Oct 2021 – Jan 2022

Kakao Enterprise

Knowledge Graph Base TF Team
  • Determined how the knowledge graph should be modeled by analyzing classes and properties within wiki data.
  • Took part in decisions on the ontology and schema used across the knowledge graph.
Apr 2021 – Oct 2021

Kakao Enterprise

Knowledge Graph AI Team
  • Extracted entities and relations from unstructured wiki text to construct RDF triples, then used SQL to profile, deduplicate, and validate the resulting data at scale.

Projects & extracurricular

Sep 2024 – Feb 2025

Hyundai Crash Engineering GenAI Project

Emory University · in collaboration with Hyundai Motor Company

Built a generative AI assistant that analyzes crash test data to support engineering decisions for vehicle safety.

Dec 2023 – Jul 2024

Korean LLM Project

Seoul National University

Developed a Korean-specialized multilingual LLM using Mixture of Experts, tokenizer expansion, instruction tuning, DPO, and ORPO; deployed publicly at dag.snu.ac.kr.

Jul 2023 – Jan 2024

Bank of Korea Financial LLM Project

Seoul National University · commissioned by the Bank of Korea

Developed a domain-specific LLM using financial raw data provided by South Korea's central bank.

Mar 2022 – Jun 2023

Medical & Legal Big Data Project

Team Lead · Korea University & Seoul National University

Led compilation of a decade of medical and legal resources, working directly with faculty from KU's Medical and Law schools.

Mar 2021 – Aug 2021

Korea University Computer Club

Mentor

Ran weekly Python training for a group of 10 students over a full semester.

Publications

2026
EACL 2026first author

CRADLE Bench: A Clinician-Annotated Benchmark for Multi-Faceted Mental Health Crisis and Safety Risk Detection

Byun, G., et al.

first author

Expert-Level Crisis Detection in Mental Health Conversations

Byun, G., Lott, A., Lipschutz, R., Minton, S. T., Stinson, E. A., & Choi, J. D.

2025
EMNLP Findings

Measuring Sycophancy of Language Models in Multi-turn Dialogues

Hong, J., Byun, G., et al.

first author

Secure Multifaceted-RAG for Enterprise: Hybrid Knowledge Retrieval with Security Filtering

Byun, G., Lee, S., Choi, N., & Choi, J.

CIKM 2025 · Applied Research

Reference-Aligned Retrieval-Augmented Question Answering over Heterogeneous Proprietary Documents

Choi, N., Byun, G., Chung, A., et al.

ACL Findingsfirst author

D-GEN: Automatic Distractor Generation and Evaluation for Reliable Assessment of Generative Models

Byun, G., & Choi, J. D.

2024
NeurIPS · Red Teaming GenAI Workshopfirst author

Large Language Model Detoxification: Data and Metric Solutions

Byun, S., & Shin, H.

EMNLP · Field Matters Workshop

The Manchu Milestone: Wav2vec 2.0 and the Future of Minority Language Speech Recognition

Seo, J., Park, M., Byun, S., et al.

LREC-COLINGfirst author

Korean Bio-Medical Corpus (KBMC) for Named Entity Recognition with Span Representation Analysis

Byun, S., et al.

LREC-COLING

KIT-19: A Comprehensive Korean Instruction Toolkit on 19 Tasks for Fine-Tuning Korean Large Language Models

Jang, D., Byun, S., Jo, H., & Shin, H.

LREC-COLING

A Study on How Attention Scores in the BERT Model Are Aware of Lexical Categories in Syntactic and Semantic Tasks on the GLUE Benchmark

Jang, D., Byun, S., & Shin, H.

LREC-COLING

ManNER & ManPOS: Pioneering NLP for Endangered Manchu Language

Lee, S., Byun, S., et al.

2023
NeurIPS · Instruction Workshopfirst author

Automatic Construction of a Korean Toxic Instruction Dataset for Ethical Tuning of Large Language Models

Byun, S., & Shin, H. — awarded NeurIPS DEI Scholarship

CARBD-Ko: A Contextually Annotated Review Benchmark Dataset for Aspect-Level Sentiment Classification in Korean

Jang, D., Seo, J., Byun, S., et al.

EMNLP · MRL Workshop

Mergen: The First Manchu–Korean Machine Translation Model Trained on Augmented Data

Seo, J., Byun, S., Kang, M., & Lee, S.

DaG LLM ver 1.0: Pioneering Instruction-Tuned Language Modeling for Korean NLP

Jang, D., Lee, S., Byun, S., et al.

Teaching

Jan 2025 – present

Emory University

Teaching assistant for Computational Linguistics (CS329), Research Practicum in Artificial Intelligence (CS371), and Introduction to Artificial Intelligence (CS211).

Nov 2023 – Aug 2024

Seoul National University

Special Lecture, Data Science Humanities

Delivered lectures on LLM training methods and practical use of the ChatGPT API to 20 undergraduates.

Sep 2023 – Feb 2024

Seoul National University

TA, Application of Computing

Assisted "Basics of Natural Language Processing"; earned a distinguished academic scholarship for the role.

Sep 2022 – Feb 2023

Seoul National University

TA, Humanities Data Science Practice

Graded assignments and developed course textbooks.

Awards & skills

Awards & scholarships

  • Excellence in Teaching Assistance, Emory2024–25
  • Academic Excellence Award, Korea University2018–21
  • Dean's List, Korea University (4.5/4.5 GPA)2020
  • ICKL 2023 Travel Award ($700)2023
  • NeurIPS DEI Scholarship ($475)2023

Skills

Languages — Korean (native), English (native), Spanish (intermediate)
Technical — Python, R, SQL, SPARQL, Praat, LaTeX