Sangyeon Yoon

Hi! I’m a Master student in Artificial Intelligence at Yonsei University, advised by Prof. Albert No.

Research Interests

I am interested in how large language models understand data and how we can make them more reliable in practice.

Robustness and evaluation: I study failure modes that are often overlooked by standard benchmarks, and develop evaluation protocols that better capture hidden behaviors, realistic risks, and model reliability in practice.
Understanding LLMs: I am interested in how language models absorb, represent, and use knowledge, and in what their reasoning processes reveal about their underlying behavior.

News

Jan 2026: Two papers were accepted to ICLR 2026. See you in Rio 🇧🇷!
Sep 2025: One paper was accepted to NeurIPS 2025. See you in San Diego 🇺🇸!
Sep 2025: Joined EXAONE Lab at LG AI Research as a Research Intern.
Aug 2025: Two papers were accepted to EMNLP 2025 Main. See you in Suzhou 🇨🇳!
Dec 2024: One paper was accepted to the NeurIPS 2024 SFLLM Workshop.

Research Experience

LG AI Research, EXAONE Lab
Research Intern, Sep 2025 – Feb 2026
Mentor: Sunkyoung Kim

Publications

(* indicates equal contribution)

Preprints

BenchPreS: A Benchmark for Context-Aware Personalized Preference Selectivity of Persistent-Memory LLMs
Sangyeon Yoon, Sunkyoung Kim, Hyesoo Hong, Wonje Jeung, Yongil Kim, Wooseok Seo, Heuiyeen Yeen, and Albert No

Position: The Term “Machine Unlearning” Is Overused in LLMs
Sangyeon Yoon*, Yeachan Jun*, and Albert No

DUSK: Do not unlearn shared knowledge
Wonje Jeung*, Sangyeon Yoon*, Hyesoo Hong*, Soeun Kim, Seungju Han, Youngjae Yu, and Albert No

Technical Reports

K-EXAONE Technical Report: Journey to Frontier-Level Performance of Foundation Models
LG AI Research
Technical Report, 2026

Peer-reviewed Conference

Rethinking Benign Relearning: Syntax as the Hidden Driver of Unlearning Failures
Sangyeon Yoon, Hyesoo Hong, Wonje Jeung, and Albert No
ICLR 2026

A2D: Any-Order, Any-Step Safety Alignment for Diffusion Language Models
Wonje Jeung*, Sangyeon Yoon*, Yoonjun Cho, Dongjae Jeon, Sangwoo Shin, Hyesoo Hong, and Albert No
ICLR 2026

R-TOFU: Unlearning in Large Reasoning Models
Sangyeon Yoon, Wonje Jeung, and Albert No
EMNLP 2025 Main

SEPS: A Separability Measure for Robust Unlearning in LLMs
Wonje Jeung*, Sangyeon Yoon*, and Albert No
EMNLP 2025 Main

SAFEPATH: Preventing Harmful Reasoning in Chain-of-Thought via Early Alignment
Wonje Jeung, Sangyeon Yoon, Minsuk Kang, and Albert No
NeurIPS 2025

Peer-reviewed Workshop

Adversarial Sample-Based Approach for Tighter Privacy Auditing in Final Model-Only Scenarios
Sangyeon Yoon*, Wonje Jeung*, and Albert No
NeurIPS 2024 SFLLM Workshop