Diffusion-based Language Generation
Exploring alternative generative paradigms for language models, including masked diffusion models and hidden-state diffusion interfaces.
Language Models · Diffusion · AI Agents · Reasoning
Injin Kong
I study how to improve language models by understanding their internal mechanisms and reasoning abilities, exploring alternative generative paradigms such as diffusion models, and developing agents that can act in real-world settings.
I am a master's student in Data Science at Seoul National University and received my B.S. in Mathematics from Seoul National University. This website collects my research, papers, education, awards, experience, and contact links.
Current Affiliation
HOLI LAB
/ Seoul National University
About
I study language models, representations, and reliable AI systems.
I am a master's student in Data Science at Seoul National University and a member of HOLI LAB. I received my B.S. in Mathematics from Seoul National University. My research focuses on large language models, representation learning, diffusion models, and interpretable AI.
I study how language models organize information internally and how their mechanisms change through training, post-training, and architectural intervention. My work connects mechanistic analysis, representation learning, diffusion-based generation, and faithful reasoning. I am also interested in agents and practical AI systems that bring research ideas into real-world applications.
Research Interests
Topics I am currently exploring.
Inner Workings of Language Models
Understanding how language models organize information internally and how their mechanisms change through training, post-training, and adaptation.
AI Agents
Building and evaluating tool-using, reliable, and autonomous systems that connect language models to real-world workflows.
Reasoning and Evaluation
Studying reasoning abilities, faithful explanations, and evaluation methods for language and multimodal models.
Representation Learning
Analyzing useful structure in high-dimensional representations, including style, values, and semantic organization in model embeddings.
Selected Papers
Research output and manuscripts.
Where Should Diffusion Enter a Language Model? Geometry-Guided Hidden-State Replacement
Injin Kong, Hyoungjoon Lee, Yohan Jo · arXiv preprint · 2026
Studies where diffusion mechanisms should enter language models through geometry-guided hidden-state replacement.
Can MLLMs Reason About Visual Persuasion? Evaluating the Efficacy and Faithfulness of Reasoning
Naeun Lee, Hyunjong Kim, Sunghwan Choi, Injin Kong, Yohan Jo · arXiv preprint · 2026
Evaluates whether multimodal large language models can reason effectively and faithfully about visual persuasion.
Mechanism Shift During Post-training from Autoregressive to Masked Diffusion Language Models
Injin Kong, Hyoungjoon Lee, Yohan Jo · arXiv preprint · 2026
Studies how model mechanisms shift during post-training from autoregressive language models to masked diffusion language models.
Dual Mechanisms of Value Expression: Intrinsic vs. Prompted Values in Large Language Models
Jongwook Han, Jongwon Lim, Injin Kong, Yohan Jo · ICML 2026
Investigates how large language models express values through intrinsic behavior and prompt-induced responses.
Style Extraction on Text Embeddings Using VAE and Parallel Dataset
Injin Kong, Shinyee Kang, Yuna Park, Sooyong Kim, Sanghyun Park · arXiv preprint · 2024
Proposes a VAE-based approach for extracting style information from text embeddings using a parallel dataset.
CV
Education, awards, and experience.
Education
2025 - Present
M.S. in Data Science
Seoul National University
2020 - 2025
B.S. in Mathematics
Seoul National University
2019
Transferred
Yonsei University
2016 - 2018
High School
Hana Academy Seoul
Awards
-
2023
Second Place
CJ Logistics Future Technology Challenge · Image-Based Volume Estimation of Parcels
-
2019
Academic Excellence Award
Yonsei University · Highest Academic Achievement
-
2018
Gold Prize
Korean Mathematical Olympiad (KMO)
-
2018
Grand Prize (1st Place)
Earth Science Competition · Hana Academy Seoul
-
2018
Gold Prize (2nd Place)
Mathematics Research Presentation Contest · Hana Academy Seoul
Experience
2023.11 - 2025.08
Chief Science Officer
Aardvark
2019
Proofreading Committee Member
Black Label Geometry
Contact
Feel free to contact me.
For research discussions, collaboration, or opportunities, feel free to reach out by email.