People › Yuchen Cao

About

Yuchen Cao is a PhD Student in the ASI Lab. of Department of Computer Science, City University of Hong Kong.

Yuchen Cao researches AI-driven software engineering, focusing on LLM-based agents for code generation, program repair, and educational feedback analysis. In SysTradeBench, he benchmarks LLM strategy-to-code trading systems under drift-aware diagnostics, finding that top models achieve ≥91.7% validity with 7.29–7.85 scores, though evidence-driven iteration induces code convergence (95.4% similarity by Iter2), suggesting LLM iteration complements rather than replaces human quant governance. His HGARena framework uses heterogeneous graph-augmented retrieval with a submitter–reviewer–executor loop, reaching a 25.3% resolve rate on SWE-bench Lite under a 20-call budget. With StuLAC, adaptive template-based caching improves LLM feedback summaries by 10.5% over manual reports and reaches 86.4% sentiment accuracy on 80,000 entries.

Research Interests

Publications (6)