M.S. student in Computer Science at the University of Chicago. I work on how LLMs reason, LLM post-training, and making what models produce auditable. Before this: a B.A. in philosophy at East China Normal University and an ML engineering internship at SHEIN.
Website · Google Scholar · LinkedIn · limengge@uchicago.edu
Looking for Summer 2027 internships in ML engineering, LLM post-training, and applied AI.
A deep research agent that traces every claim to its source. Crash-safe runs, a judge that checks each citation, and follow-up questions that reuse the evidence.Python · FastAPI · LangGraph · Neo4j · React Demo · Code |
Code and benchmark for my AAAI 2026 paper: 11,000 syllogisms show that five frontier LLMs fail in the same places, for the same reasons.Python · asyncio · WordNet Paper · Code · Try the game |
- Consistent Biases in Large Language Models' Syllogistic Reasoning. AAAI 2026 Bridge on Logical and Symbolic Reasoning in Language Models. [paper]
- From Output Quality to Auditability: A Data Mining Agenda for LLM-Generated Artifacts. With Mingjia Qian. ICDM 2026 BlueSky Track (accepted).
More on my website.
