About

I am a 2nd-year Ph.D. student in Computer Science at Johns Hopkins University, advised by Tianmin Shu. I’m also a research intern at Meta FAIR, mentored by Jason Weston.

I received bachelor degrees in Honors Computer Science and Mathematics from New York University (2020-2024), where I was awarded the Computer Science Prize for the Most Promising Student. I was as a research intern at MIT (2023), advised by Josh Tenenbaum. I received an Outstanding Paper Award at ACL 2024 for my work on multimodal Theory of Mind.

I’m interested in developing AI systems with advanced social intelligence (ASI). My current areas of focus include:

Reasoning about Human Interaction: Developing AI that can continuously perceive, reason about, and respond to human behavior and cognition.

• We all want our AI to understand us. My research addresses this through a model-based approach, developing agent models that capture human behavior and infer mental states (e.g., goals, beliefs, desires) from observations (e.g., speech, motion).
• In particular, my works automate agent modeling with LLM backends, advancing Theory of Mind reasoning by grounding it in cognitive theories, learning from multi-modal, multi-agent and multi-perspective observations, and scaling to long-term and complex settings.

Learning from Human Interaction: Building AI that continuously learns from real-world human interactions.

• We advocate for an era of real-world human interaction that learns from real users' diverse goals and preferences, adapts to situational demands, and scales with actual usage. We introduce Reinforcement Learning from Human Interaction (RLHI), which learns directly from user conversations and outperforms RLHF at the user level.
• I initiated RSS Workshop on Continual Robot Learning from Humans.

Besides academics, I enjoy playing Go, piano, badminton, hiking, and so much more.

News

• [Oct, 2025] Honored to receive the Amazon PhD Fellowship!
• [Oct, 2025] Invited talk @ Google on RLHI.
• [Sep, 2025] AutoToM is accepted by NeurIPS 2025 as a Spotlight Presentation!
• [June, 2025] Co-organized the Workshop on Continual Robot Learning from Humans at RSS 2025.
• [June, 2025] I was selected as a Notable Reviwer for ICLR 2025.
• [Jan, 2025] MuMA-ToM is accepted by AAAI 2025 as an Oral Presentation!
• [Sep, 2024] Invited talk @ University of Washinton on “Towards Multimodal Theory of Mind”.
• [Aug, 2024] MMToM-QA won the Outstanding Paper Award at ACL 2024! It was also covered by Futurity, Synced, and JHU News.

Recent Publications

(* denotes equal contribution; † denotes project lead)
For a more complete list, please see my publication page or Google Scholar page.

The Era of Real-World Human Interaction: RL from User Conversations Chuanyang Jin, Jing Xu, Bo Liu, Leitian Tao, Olga Golovneva, Tianmin Shu, Wenting Zhao, Xian Li, Jason Weston
arXiv Preprint (DAIR.AI Weekly Papers Top-3; Invited Talk at Google)
paper / tweet Reinforcement Learning from Human Interaction (RLHI) learns directly from real-world user conversations, beating RLHF at the user level and enabling personalized, contextual, and continually improving AI assistants.

AutoToM: Scaling Model-based Mental Inference via Automated Agent Modeling Zhining Zhang*, Chuanyang Jin*†, Mung Yao Jia*, Shunchi Zhang*, Tianmin Shu (†: project lead)
NeurIPS 2025 (Spotlight)
paper / project / code AutoToM is an automated agent modeling method for scalable, robust, and interpretable mental inference. It achieves SOTA on five benchmarks, produces human-like confidence estimates, and supports embodied decision-making.

OS-Genesis: Automating GUI Agent Trajectory Construction via Reverse Task Synthesis Qiushi Sun*, Kanzhi Cheng*, Zichen Ding*, Chuanyang Jin*, Yian Wang, Fangzhi Xu, Zhenyu Wu, Chengyou Jia, Liheng Chen, Zhoumianze Liu, Ben Kao, Guohao Li, Junxian He, Yu Qiao, Zhiyong Wu
ACL 2025 (Huggingface Daily Papers Top-1)
paper / project / code / model / data / slides OS-Genesis enables agents to actively explore web and mobile environments, facilitating the self-collection and self-annotation of high-quality training data.

MuMA-ToM: Multi-modal Multi-Agent Theory of Mind Haojun Shi*, Suyu Ye*, Xinyu Fang, Chuanyang Jin, Leyla Isik, Yen-Ling Kuo, Tianmin Shu
AAAI 2025 (Oral)
paper / project / code / benchmark / leaderboard / slides MuMA-ToM introduces a benchmark and method that extend MMToM-QA to multi-agent interaction scenarios.

MMToM-QA: Multimodal Theory of Mind Question Answering Chuanyang Jin, Yutong Wu, Jing Cao, Jiannan Xiang, Yen-Ling Kuo, Zhiting Hu, Tomer Ullman, Antonio Torralba, Joshua Tenenbaum, Tianmin Shu
ACL 2024 (Outstanding Paper Award)
paper / project / code / benchmark / leaderboard / slides / Futurity news / JHU news
Can machines understand people's minds from multimodal inputs? MMToM-QA introduces the first benchmark and a promising method that combines the flexibility of LLMs with the robustness of Bayesian inverse planning.

Feel free to check out my undergrad projects. A mountain of gratitude to those who have kindly mentored and inspired me with their vision and passion!

Selected Honors & Awards

• Amazon PhD Fellowship, 2025
• Notable Reviewer, ICLR 2025
• Outstanding Paper Award, ACL 2024
• Presidential Honors Scholar and Summa cum Laude, New York University (Top 5%), 2024
• Computer Science Prize for the Most Promising Student, New York University (1 person/year), 2023
• Dean’s Undergraduate Research Fund, New York University, 2023
• COMAP International Scholarship Award (Top 0.1%), 2022
• MAA Award in Mathematical Contest in Modeling (Top 0.1%), 2022
• Bronze Medal of Shing-Tung Yau Computer Science Award (Top 1%), 2019
• Finalist of FIRST Robotics Competition World Championship (Top 0.2%), 2019
• NFLS Outstanding Student Leader Award and Zhou Enlai Scholarship (Top 1%), 2018
• First Prize of Chinese Mathematical Olympiad (Top 0.1%), 2018
• Champion of International Regions Mathematics League, 2018