Penghao Jiang

Vul data with RL

SkillsBench evaluates how well skills work and how effective agents are at using them

A modular framework for code translation with proactive semantic guidance. It combines semantic summaries and static analysis to infer pitfalls before translating. Features an automated test-driven repair loop with early-exit and diagnostic feedback. Organized into distinct agents (semantic, inference, execution) to facilitate ablation studies.

PyCTrace – Python-C Cross-Language Call Graph Error Detection Tool