- Data/: This stores data cleaned data of coontext, human_response and AI response. 'cleaned_equal.csv' has reponses where AI and human reponse match. 'cleaned_non_equal.csv' has reponses where AI and human reponse doesn't match.
- preview_data: This is just to view data in text format rather than dataframe.
- preview_data/: This is just to view data in text format rather than dataframe.
- training_data/: This folder stores data X_train(tokenised format of context/response/agent_response) and y_train: LLM generated score against them.
- scoring_responses/: This directory has files to evaluate and score AI responses -
- format_prompts.py: Python file to take random data and generate prompt (to be used in ChatGPT) for the same.
- prompts/: saves prompts used for GPT3.5 to generate scores for unlabeled data.
- scrape_chat_gpt_data.py: Python file to scrape data from ChatGPT chat, saves to 'training_data/'.
- training_scoring_model.py: This file uses data to train our scoring model and saves to 'model/'.
- scoring_responses.py: This file is used to generate scores for responses.
</ul>