Shinwoo Park

CV

Name: Shinwoo Park

Degree: Ph.D. in Artificial Intelligence

Affiliation: Yonsei University, Seoul, South Korea

Graduation Date: February 2026

Research Interests

My research focuses on ensuring the safety, transparency, and accountability of large language models (LLMs) through detection and watermarking techniques. I develop multi-modal and multi-lingual systems that identify or trace LLM-generated content across natural language and source code domains.

Specifically, I explore two complementary directions: (1) linguistic and stylistic feature–based detection, which analyzes morphological, syntactic, and stylistic patterns to distinguish human- and LLM-generated text or code; and (2) LLM watermarking, which embeds imperceptible yet verifiable statistical or structural signals into generated outputs.

My recent works include KatFishNet, the first linguistic feature–based detector for Korean text; LPcodedec, a coding-style-driven detector for paraphrased code; STELA, a syntactic-predictability watermark enabling model-free detection; and WaterMod, a probability-balanced modular watermarking framework supporting multi-bit payloads.

Broadly, my goal is to build trustworthy generative systems that are interpretable, regulation-compliant, and resistant to misuse.

Research Summary

My research aims to promote responsible and verifiable AI generation by developing reliable methods for detecting and attributing LLM-generated text and code. I pursue two main directions that mutually reinforce each other:

• Linguistic/Stylistic Feature-based Detection: Models such as KatFishNet and LPcodedec analyze linguistic or coding-style cues—word spacing, part-of-speech diversity, punctuation patterns, naming and indentation consistency—to capture distributional differences between human and LLM authors.
• LLM Watermarking: Frameworks such as STELA and WaterMod embed imperceptible signals during generation using linguistically or probabilistically adaptive mechanisms, enabling publicly verifiable and multi-bit attribution without harming fluency.

These systems demonstrate strong multilingual (English, Korean) and multimodal (text + code) generalization, advancing interpretable and regulation-aligned AI provenance research.

Research Statement

My long-term research vision is to establish a unified framework for provenance-aware and interpretable AI that spans both language and programming modalities. To achieve this, I combine linguistic insight, statistical modeling, and watermark design to construct transparent interfaces between human communication and generative models.

Linguistic / Stylistic Feature-based Detection:
My work on KatFishNet introduces the first benchmark and detector for LLM-generated Korean text, leveraging word-spacing irregularities, POS n-gram diversity, and comma usage to expose cross-morphological differences between human and machine writing. Extending this idea to source code, LPcodedec identifies LLM-paraphrased code by quantifying coding-style features such as naming consistency, indentation regularity, and comment ratio.

LLM Watermarking:
My research advances from distribution-based watermarking to linguistically adaptive and probability-balanced methods. STELA modulates watermark strength according to syntactic predictability modeled by POS n-gram entropy, enabling model-free public detection. WaterMod generalizes this concept through modular token-rank partitioning that guarantees at least one high-probability token per class, supporting zero-bit and multi-bit watermarking with minimal quality loss.

Together, these studies form a coherent agenda for trustworthy and interpretable generative AI, bridging linguistic analysis and information-theoretic watermark design to meet emerging transparency and safety requirements.

Publications

† First author    †* Co-first author (marked with *)

To Appear / Published

Under Review

Projects

• Topic Modeling for Entity Retrieval (2021–2022, Funded by Ministry of Science and ICT, Korea):
  Developed a knowledge-graph entity search module from unstructured text using LDA-based topic modeling. Designed a query-aware retrieval mechanism to align user queries with semantically relevant entities.
• Medical Text Mining (2022–2024, Industry-funded by Soldoc):
  Built an NLP-based clinical support framework from psychiatrist–patient dialogues. Applied BERT embedding clustering to discover latent topics and extract suspected mental health conditions with interpretable evidence keywords from conversational data.
• Human-AI Programming Lab (2023–2025, Funded by National Research Foundation of Korea):
  Conducted research on AI techniques for collaborative programming systems, including code search, code question answering, worst-case time complexity prediction, and detection of LLM-generated code.
  Publications:
  • EACL 2023 (Main Track, First Author): Natural language-based code search and code QA for developer assistance.
  • Engineering Applications of Artificial Intelligence (EAAI), 2025 (First Author, 2 papers): (1) Contrastive learning-based worst-case time complexity prediction for source code, (2) Detection of LLM-generated code using coding-style features.
  • Expert Systems with Applications (ESWA), 2025 (Co-author): In-context learning approach for worst-case time complexity prediction.
• Research on Effective Watermarking Techniques for AI-generated Codes (2025, Funded by National Research Foundation of Korea):
  Investigated watermarking methods for LLM-generated code by embedding identifiable signals during the generation process while preserving functionality and code quality.
  Publication:
  • Findings of EACL 2026 (Co-first Author): A watermarking framework for AI-generated code enabling reliable attribution while maintaining syntactic and semantic integrity.

Professional Services

• Reviewer, ACL Rolling Review (ARR) (2023--)
  
• Reviewer, ACL Student Research Workshop (SRW) (2026--)
  Contributed to mentoring and evaluating early-stage research in NLP.

Skills

• Programming: Python, Bash
• ML Frameworks: PyTorch, scikit-learn, Hugging Face Transformers
• NLP: SpaCy, NLTK, KoNLPy, KiwiPiePy
• LLM APIs: OpenAI API, Gemini API, Ollama
• Data Analysis: Pandas, NumPy, SciPy
• Visualization: Matplotlib, Seaborn
• Version Control: Git
• Writing: LaTeX
• Languages: Korean (native), English (fluent)