AutoAC

Agentic Self-Refinement for Competitive Programming

OSU SP25 CSE 5525 Final Project.

Installation:

1. Install conda if you haven't already from the installation page.
2. In the project's home directory:

conda create --name autoac python=3.10

3. Install pip requirements

pip install -r requirements.txt

4. Download a local LLM or use an API key, making changes to the agents/model_api_mappings.json file as necessary.We'd recommend downloading Meta's Llama-3.1-8B-Instruct model from the Huggingface hub to local-models/.
5. Run the demo story completion with

python3 framework_usage.py

Downloading Data:

We use the CSES.fi dataset curated by juvi21 on HuggingFace. Download this data using

cd dataset
python3 download_data.py