Turning raw data into cozy little stories.

> UIUC MSIM student · Data Enthusiast · Software Developer

Python Altair Vega-Lite HTML/CSS
Portrait of Zhongyin Wang

A Bit About Me

From transportation engineering to data storytelling.

I'm an MSIM student at UIUC focused on software engineering, databases, and data visualization — currently building things like Nexus, an AI-powered academic planning system, and a graph-based compliance platform with Neo4j. Before that: civil, hydraulic & transportation engineering at Zhejiang University, which is where the analytical habit started.

Python Java Neo4j LLMs
More about me →

Currently

M.S. in Information Management

University of Illinois Urbana–Champaign · 2025–2027

Research

Research on making LLM systems more reliable.

See all research →
Main figure from “Truncated Reasoning Under a Budget: Evaluating System-1 Models as Decision-Makers” First author

Under review, ACL Rolling Review · Oct 2026

Truncated Reasoning Under a Budget: Evaluating System-1 Models as Decision-Makers

When a reasoning model is cut off mid-thought, a System-1 decision model can read the partial trace and answer. At a 1,500-token budget it beats forced answering by 4.1 pp, and a learned stopping policy saves 71.3% of the full-run cost for a 0.58 pp accuracy loss.

LLM ReasoningEfficiencyEvaluation
Main figure from “Decision Models as Judges for Hallucination Detection”

Under review, ACL Rolling Review · Oct 2026

Decision Models as Judges for Hallucination Detection

Can cheap decision models stand in for LLM judges? Across five hallucination types (9,852 items), they match or come close on some types, and a confidence cascade reaches a generative judge's accuracy at 1–3% of its cost — but needs validation before use.

LLM-as-a-JudgeHallucinationCalibration

Projects

A couple of builds worth a second look.

See all projects →

Got a dataset that needs a story?

Always happy to talk data, viz, or the occasional Bigfoot sighting.