Research Assistant · UIUCU Lab / ATLAS Moonshot
Researching how capable agents become systems that can work, learn, and persist together.
I’m a researcher at the University of Illinois Urbana-Champaign working on language-model agents, multi-agent systems and human–agent interaction. I work with Jiaxuan You, Jiawei Han and ChengXiang Zhai across agent organizations, evidence-grounded evaluation and graph-grounded AI.
My central question is how useful capabilities can become properties of a system—not just of the model or the individual agent inside it. My work spans mathematical modeling, communication protocols, long-running adaptation, organizational memory and the interfaces through which people participate.
RELIC is the current focus of my organization-level research. Alongside it, RINI studies unsupported novelty claims and evidence retrieval in automated research, while my work in the ATLAS Moonshot connects structured skill graphs to personalization and learner modeling.
My work spans problem formulation, professional-profile graphs, the OrgEnv simulation harness and evaluation of product-development activity. RELIC studies how shared state, governed protocols and reusable capabilities persist across work and member turnover; its Human–Agent Organization extension connects human participation to shared review and governance.
Explore RELICLeading the personalization and learner-modeling direction: Open Reskill Graph structures, prerequisite relations, uncertainty-aware learner states and evaluation for trustworthy AI-assisted learning and reskilling.
ATLAS announcementProblem formulation, novelty taxonomy, evidence alignment and evaluation of generated research content. This direction now connects to RINI’s study of how missing evidence can be mistaken for originality.
Explore RINILed benchmark and scenario design, protocol-behavior analysis, and implementation of scenario-aware protocol selection. Published at ICML 2026.
Explore ProtocolBenchDesigned and implemented the Werewolf scenario and analysis pipeline to study theory of mind, cooperation, leadership and distrust in decentralized agent interactions. Published at ACL 2025.
Explore MultiAgentBenchGPA 3.79 / 4.00. Graduate-level Advanced Natural Language Processing: A+, 100/100.
My training also includes computational photography, systems programming, algorithms, numerical methods and computer architecture.
My research connects work with Jiaxuan You, Jiawei Han and ChengXiang Zhai at UIUC. I also began a remote collaboration with Manling Li’s group at Northwestern in October 2025, working with Qineng Wang on language-conditioned 3D affordance prediction.
Northwestern collaboration
University of Illinois Urbana-Champaign
Questions and perspectives on this page are welcome.
This conversation belongs to About. Comments appear only after review. For contact details or personal matters, use the private message form.
Loading comments…