Jiayi Wu (jyah-yee/jiā-yì)

Hello:)

Hi! I'm an undergraduate studying Mathematics-Computer Science (Sc.B.) and Applied Mathematics (A.B.) at Brown University.

I hope to build the formal and empirical infrastructure for trustworthy machine learning across two complementary dimensions:

On the construction side, model architectures informed by various learning paradigms¹ (neurosymbolic programming² in particular), and how they enable more robust and interpretable reasoning models.
On the guarantee side, evaluation and auditing pipeline design³ informed by formal verification^{5, 1}, and how they scaffold more aligned and accountable automated systems.

Concretely, I've been interested in formal theorem proving and program synthesis as fully specified, unambiguous, machine-checkable substrates for studying models' reasoning processes; I'm also interested in how they interact and integrate with general-purpose foundation-model systems.

My Work

Below are selected projects and publications; a full list of research projects is available in my academic CV.

Pattern Mining and Automated Tactic Discovery in Lean 4 Theorem Proving, in collaboration with Gavin Zhao et al., advised by Robert Lewis and Stephen Bach
Formalizing and Benchmarking Overton Pluralism in LLMs [Preprint], in collaboration with Elinor Poole-Dayan et al., advised by Michiel Bakker, accepted to the Fourteenth International Conference on Learning Representations (ICLR 2026) and NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle: Benchmarks, Emergent Abilities, and Scaling
Motif of Thoughts (MoT): Reusable Abstractions for Compositional Neurosymbolic Reasoning, in collaboration with Chance Jiajie Li et al.
- Simulating Society Requires Simulating Thought (position paper on causally grounded reasoning modules for multi-agent systems) [Preprint], accepted to the Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS 2025)
- HugAgent: Evaluating LLMs in Simulating Individual Reasoning on Open-Ended Tasks [Preprint] [Repository], accepted to NeurIPS 2025 Workshop on LAW 2025: Bridging Language, Agent, and World Models (Spotlight), PersonaLLM: Workshop on LLM Persona Modeling (Oral), and Socially Responsible and Trustworthy Foundation Models (ResponsibleFM)
Pluggable Analyses for Modern Real Systems, Undergraduate Teaching and Research Awards (UTRA) project in collaboration with Lukas Lazarek et al., advised by Nikos Vasilakis
Algorithmic Fairness and AI Governance Primer at Socially Responsible Computing (SRC) Handbook [Handbook], advised by Suresh Venkatasubramanian and Julia Netter

Misc

Outside of research, I help organize communities at Brown that connect technology with governance and public-interest perspectives. I’m co-president of Brown’s AI and Robotics Ethics Society (AIRES) and an undergraduate teaching assistant at CSCI 1952B Responsible Computer Science in Practice; I also co-directed the AI Governance Panel at Brown China Summit 2025; more updates are available here!

In my spare time, I enjoy cold brew, taking on plank challenges (≥7min), and roaming around the neighborhoods and community spaces I'm involved in 🏘️🌳 -- if I'm not in CIT^;