01
Trustworthy LLMs
Detecting and reducing hallucination, grounding reasoning in evidence, and building evaluations that hold up.
Look Before You LeapEMNLP 2026 Findings
Aligning with Your Own VoiceACL 2026 Findings
Chung-Ang University, Seoul
Intelligent Information Processing Lab
중앙대학교 지능형 정보처리 연구실
We study how large language and multimodal models reason, where they fail, and how to make them trustworthy, safe, and efficient.
90Publications
22Members
9Active projects
Recent acceptances
Research
Language and multimodal AI that people can rely on
Our current work clusters into six directions. Each card lists recent papers as a starting point.
02
Safety and Unlearning
Safety alignment for reasoning models, selective safety steering, and machine unlearning that stays robust under bias.
Beyond Safe AnswersEMNLP 2026
Easy to Learn, yet Hard to ForgetAAAI 2026
03
Fairness and Multilingual NLP
Measuring and mitigating gender, dialect, and cultural bias, with a focus on Korean language technology.
FairQEACL 2026
Steering LLMs toward Korean Local SpeechLREC 2026
04
Multimodal AI
Vision-language models that are efficient and reliable, from visual token pruning to OOD detection and imaging.
RI-PruneCIKM 2026 Oral
HyCalCVPR 2026
05
Efficient Reasoning and Retrieval
Prompt compression, multi-hop retrieval, and small language models that reason well at low cost.
IterCOMPACL 2026
Beyond One-Shot ExpansionCIKM 2026
06
AI for Research
How LLMs can responsibly support peer review and citation verification.
LLMs in the Peer Review ProcessEACL 2026 Findings
RefLensAAAI 2026 Demo
News
What's new
Papers, awards, and grants from the past three years, newest first.
Lab life
At conferences, together
Join us
함께 연구할 학생을 찾습니다
자연어처리와 멀티모달 AI를 함께 연구할 학부 연구생과 대학원생(석사, 박사)을 모집합니다. 관심 있는 분은 간단한 자기소개와 이력서를 보내 주세요.
지원 메일 보내기
ybkim85@cau.ac.kr
TRACK 1
일반대학원 AI학과
LLM 기반 언어 이해와 추론, 편향과 안전성 분석, 다국어 모델링, LLM 에이전트 기반 평가와 검증, 구조적 검색과 추론
TRACK 2
첨단영상대학원 영상학과
비전 언어 표현 학습, 텍스트 이미지 융합 모델, 약지도 이미지 해석, 도메인 일반화, 멀티모달 균형 조정, 의료와 과학 데이터 응용