About Me
Hi! I am a research engineer on the Foundation Research Team at NAVER Cloud.
Previously, I led the Application LM Team (Alignment Team) at NCSOFT.
My current research focuses on alignment-tuning and LLM evaluators.
Research Experience
- LLM-as-a-judge: OffsetBias: Leveraging Debiased Data for Tuning Evaluators
- VARCO LLM: Alignment Tuning of Foundation Model
- VARCO Text: NCSOFT’s AI Writing Assistant
- Weather Forecast News Article Generation (with Yonhap News Agency)
- Baseball Game Summary Generation from the Game Log
- Deep Research on Pointer-Generator(Copy) Network Reducing Hallucinations
- Developing In-house Generative AI-model-training Framework
- Developing Korean Word2vec Model
For more details, please visit my Linked-in profile. Click here
Educations & Careers
NAVER Cloud
- Research Engineer (2024.09~)
Global AI Platform Corporation - Machine Learning Engineer (2024.08-2024.09)
NCSOFT NLP Center
- Application LM Team Lead (2023.09-2024.08)
- Generation Model Team Lead (2022.06-2024.01)
- Research Engineer (2018.01-2024.08)
Seoul National University
- M.S. Student in Computer Science and Engineering (2015.03-2017.02)
Stayes Inc.
- CTO, Co-founder (2014.02-2015.02)
Visionaries Inc.
- CTO, Co-founder (2013.05-2014.02)
Korea University
- B.S. Student in Computer Science and Engineering (2011.03-2015.02)
Publications
-
Junsoo Park, Seungyeon Jwa, Meiying Ren, Daeyoung Kim, Sanghyuk Choi. OffsetBias: Leveraging Debiased Data for Tuning Evaluators. EMNLP2024 Findings, 2024.
-
Sanghyuk Choi, Jeong-in Hwang, Hyungjong Noh, Yeonsoo Lee. May the Force Be with Your Copy Mechanism: Enhanced Supervised-Copy Method for Natural Language Generation. arXiv preprint, 2021.
-
Sanghyuk Choi, Taeuk Kim, Jinseok Seol, Sang-goo Lee. A Syllable-based Technique for Word Embeddings of Korean Words. SCLeM2017(The 1st Workshop on Subword and Character level models in NLP, EMNLP2017).
-
Sanghyuk Choi, Jinseok Seol, Sang-goo Lee. On Word Embedding Models and Parameters Optimized for Korean. The 28th Annual Conference on Human & Cognitive Language Technology.
-
Sanghyuk Choi, Sang-goo Lee. On the Implementation of an Input and Management System for Korean Minsokak(Folk Music) Fieldwork Material Database. Studies in Korean Music(Vol.59).
Open Source
- Contributor of OpenNMT-py: Open-Source Neural Machine Translation and (Large) Language Models
- Contributor of PEFT: State-of-the-art Parameter-Efficient Fine-Tuning (PEFT) methods, Huggingface
AWARDS
| Year | Award | Category |
|---|---|---|
| 2013 | 9th Prize | ACM-ICPC Asia Daejeon Regional |
| 2012 | 4th Prize | ACM-ICPC Asia Daejeon Regional |
| 2012 | 2nd Prize | Software Olympiad, Korea University |