Jeewoo Sul

👋 Hello, I am Jeewoo Sul (설지우).

I am an AI Research Engineer at LG Electronics’ AI Lab (Seoul, Korea) since February 2024. My goal is to build AI systems that are reliable and beneficial. I am currently developing AI Coding Agent using large language models (LLMs). My work explores how AI can support and optimize every stage of the software development lifecycle.

🔍 Current Research Interests

LLM-as-a-Judge: Utilizing large language models for robust evaluation of code quality and model performance.
Prompt Engineering: Designing and optimizing prompts (e.g., Persona, CoT) to achieve high-performance outcomes in software development features.
Reflection Agent: Exploring an adaptive multi-agent framework employing iterative self-reflection, verbal reinforcement, and collaborative simulation.

🔙 Past Research Experience
As a graduate student majoring in Computer Science at Hanyang University, I conducted research in Natural Language Processing under the supervision of Prof. Yong Suk Choi, focusing on Abstractive Text Summarization and Prompt Tuning.

🤗 I’m excited about potential collaborations—please don’t hesitate to get in touch!
🚨 Note: Due to the rise in phishing emails, I kindly request that you contact me via LinkedIn or a verified email address. Emails from unverified sources may not receive a response. Thank you for your understanding!

news

Jun 18, 2024	I attend NAACL 2024 at Mexico City, Mexico and present our poster!
Feb 13, 2024	I’m starting a new position as a research engineer at LG Electronics!
Jul 15, 2023	I attend ACL 2023 at Toronto, Canada and do an oral presentation (Session: Summarization).

selected publications

NAACL short

Is Prompt Transfer Always Effective? An Empirical Study of Prompt Transfer for Question Answering

Minji Jung^*, Soyeon Park^*, Jeewoo Sul^*, and 1 more author

In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers), Jun 2024

Abs DOI Code

Prompt tuning, which freezes all parameters of a pre-trained model and only trains a soft prompt, has emerged as a parameter-efficient approach. For the reason that the prompt initialization becomes sensitive when the model size is small, the prompt transfer that uses the trained prompt as an initialization for the target task has recently been introduced. Since previous works have compared tasks in large categories (e.g., summarization, sentiment analysis), the factors that influence prompt transfer have not been sufficiently explored. In this paper, we characterize the question answering task based on features such as answer format and empirically investigate the transferability of soft prompts for the first time. We analyze the impact of initialization during prompt transfer and find that the train dataset size of source and target tasks have the influence significantly. Furthermore, we propose a novel approach for measuring catastrophic forgetting and investigate how it occurs in terms of the amount of evidence. Our findings can help deeply understand transfer learning in prompt tuning.
ACL short

Balancing Lexical and Semantic Quality in Abstractive Summarization

Jeewoo Sul, and Yong Suk Choi

In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), Jul 2023

Abs DOI Code

An important problem of the sequence-to-sequence neural models widely used in abstractive summarization is exposure bias. To alleviate this problem, re-ranking systems have been applied in recent years. Despite some performance improvements, this approach remains underexplored. Previous works have mostly specified the rank through the ROUGE score and aligned candidate summaries, but there can be quite a large gap between the lexical overlap metric and semantic similarity. In this paper, we propose a novel training method in which a re-ranker balances the lexical and semantic quality. We further newly define false positives in ranking and present a strategy to reduce their influence. Experiments on the CNN/DailyMail and XSum datasets show that our method can estimate the meaning of summaries without seriously degrading the lexical aspect. More specifically, it achieves an 89.67 BERTScore on the CNN/DailyMail dataset, reaching new state-of-the-art performance. Our code is publicly available at \urlhttps://github.com/jeewoo1025/BalSum.