Zichen Wang

Can an LLM provide a low-cost, scalable measure of Big Five traits from student letters, and do these measures meaningfully predict future outcomes? We treat the LLM as an additional informant in a multi-informant framework and show that AI ratings retain distinct predictive power alongside teacher, student, and guardian reports.

Using ~541,000 arXiv CS preprints linked to OpenAlex and a continuous DID design around the August 2018 NIH investigations, we find that topics with greater overseas-Chinese diaspora exposure experienced a persistent citation decline — while publication volume, content novelty, and first-year citations remained stable, pointing to weakened long-run knowledge diffusion.

An AI-native paper reading and research platform. The web reader supports agent-generated deep-reading reports that help researchers filter papers worth a careful read and understand them faster. Paper APIs let local agents (Codex, Claude Code) plug into a personal paper library for literature management and cross-paper work.

A "cyber mentor" for junior researchers — undergraduates, graduate students, and early-career faculty — built on a structured knowledge base of ~40,000 papers from 30+ top economics and business journals (2000–2026). Using agentic search over this corpus, it supports academic writing, econometric method learning, research topic discovery, and ongoing research guidance.