Resume
Research Experience
- Developed the AutoCoder model for code generation tasks, outperforming GPT-4 Turbo (April 2024) and GPT-4 on the HumanEval benchmark dataset.
- Involved in the MACM project, which utilizes a multi-agent system for condition mining to solve complex mathematical problems.
- Created InfantAgent, a multimodal agent capable of interacting with its PC in a multimodal manner.
- Developed GUI_Spotlight, a think-with-image GUI visual grounding model.
- Contributed to the HPC-Fortran2CPP/OpenMP-Fortran-CPP-Translation project, supporting translation between OpenMP Fortran and C++ code.
- Participated in the Fortran2Cpp project, focusing on Fortran to C++ code conversion.