Ke Zhang photo

Ke Zhang (张轲)

PhD Student in Mathematics • University of California, Riverside

I build tool-using AI agents for domains where outputs can be checked: Lean, scientific simulators, and optimization solvers. My research asks when these agents’ results can actually be trusted.


Selected Research

Compile rate versus faithfulness across eight systems
Ke Zhang, Patricio Gallardo, Sudhir Murthy, Yi Xie, Zhi Wang, Maziar Raissi
arXiv preprint, 2026
Accepted at the MATH-AI Workshop @ NeurIPS 2026 and the NeurIPS 2025 LLM Evaluation Workshop (earlier version: Paper · Poster).
Same simulation, different evidence: with the report inline the agent reads it and answers correctly; with only an index it reads its own file, truncates the value, and answers wrongly. Below, the three-step audit: build, exposure, closure.
Ke Zhang, Sahchit Chundur, Jerry Peprah Owusu, Mohammad Javad Abdolhosseini Qomi, Maziar Raissi
arXiv preprint, 2026
PHREEQC agent workflow
Ke Zhang, Sahchit Chundur, Mohammad Javad Abdolhosseini Qomi, Maziar Raissi
arXiv preprint, 2026
The ToolGate generation and acceptance loop
Ke Zhang, Yankang Liu, Roya Zandi, Maziar Raissi
arXiv preprint, 2026
Viral capsids as curved elastic shells
Yankang Liu, Ke Zhang, Maziar Raissi, Roya Zandi
AI&PDE Workshop @ ICLR 2026 — Poster
Agentic tool-use workflow
Ke Zhang, Maziar Raissi
Workshop paper, 2025 · AAAI 2026 Workshop (AI4Research) — Accepted

About Me

Outside research, I climb mountains — Mt. Siguniang Erfeng (5,276 m / 17,309 ft) in Sichuan, and Mt. San Jacinto via Devil's Slide Trail (16 mi) — and I do my own car maintenance, down to the oil, brake pads, and rotors.


Contact

Email: kzhan153@ucr.edu