Researcher at Zhejiang University working on LLM agents, with a focus on automated systems that can configure, run, and evaluate software environments.
Dongdong Hua is a researcher at Zhejiang University. His current interests center on LLM agents and practical automation for real software systems.
The work explores how language-model agents can reason across repositories, dependencies, tools, and execution feedback to make heterogeneous codebases runnable.
RAT proposes a language-agnostic framework for automated environment configuration on arbitrary repositories. The paper also introduces RATBench, a benchmark for evaluating environment setup across heterogeneous real-world repositories.