Minwoo (Josh) Kang
I work on LLM post-training at Mistral. Previously, I received my Ph.D. in Computer Science at UC Berkeley and Berkeley AI Research (BAIR), advised by John Canny and John Wawrzynek.
My academic research has focused on various topics in language modeling, NLP, and beyond:
- User Simulation via Language Models
- Conversational and Collaborative AI Agents
- Agentic AI for digital hardware design, optimization, and verification
For contact, please reach out to:
mkang at cs dot {my grad school institution} dot edu
news
| Oct 06, 2026 | Mistral Large 4 is now available in public preview! |
|---|---|
| Sep 29, 2025 |  I will be attending COLM 2025 to present our work “Deep Binding of Language Model Virtual Personas: a Study on Approximating Political Partisan Misperceptions”! I’m also co-organizing PragLM 2025 Workshop (October 10th) co-located at COLM! Looking forward to amazing meetings and the workshop! |
| Feb 28, 2025 | Check out our new preprint “Language Model Fine-Tuning on Scaled Survey Data for Predicting Distributions of Public Opinions”! Link to X thread →  . |
| Nov 08, 2024 | I will be attending EMNLP 2024 to present our work “Virtual Personas for Language Models via an Anthology of Backstories”. Looking forward to presenting our poster! |
selected publications
-
- Quantifying the Utility of User Simulators for Building Collaborative LLM AssistantsarXiv preprint, 2026
- Building Interactive Real-Time Agents with Asynchronous I/O and Speculative Tool CallingarXiv preprint, 2026
- Identity, Cooperation and Framing Effects within Groups of Real and Simulated HumansarXiv preprint, 2026