Replication package for Automated Extract Method Refactoring with Open-Source LLMs

Comparison of LLMs for Automated Code RefactoringThis project compares various Large Language Models (LLMs) for the task of automated code refactoring. The framework allows you to test and evaluate multiple models on real-world Python code samples.---Features- Supports 5 popular LLMs for code refactoring- Works on sample Python files in batch- Flexible model configuration via `model.yaml`- Designed for experimentation and evaluation in research or production- Uses CodeNet datasetSpecify one of the following in `CodeBase/model.yaml`:- `Qwen/CodeQwen1.5-7B-Chat`- `deepseek-ai/deepseek-coder-6.7b-instruct`- `meta-llama/Llama-3.2-3B-Instruct`- `Qwen/Qwen2.5-Coder-7B-Instruct`- `microsoft/Phi-4-mini-instruct`Specify one of the following in `CodeBase/prompts.yaml`:- `few_shot`- `zero_shot`- `rci`Dependencies:- pip install -r requirements.txtData:- data/Python_wrapped - data/Problem_descriptions - selected_files.txtRun :- inference.main