Leveraging Large Language Models for Concept Graph Recovery and Question Answering in NLP Education

This is the GitHub repo for the paper submission.

src: Zero-shot and Ablation study code. Run with >> python xxxx.py. We include LLaMa, GPT code.

simple_baselines_src: Simple classifier baseline source code.

supervised_baselines_src: GCN-based models for supervised methods. Check the readme for more information.

embedding_src: code and embedding for applying LLaMa and GPT models.

RAG_src: Source code for RAG settings.

concept_data: dataset containing 3 domains.

data_generator: help code for making and cleaning TutorQA benchmark.

data_generator/TutorQA: the final TutorQA benchmark. We also include LLaMa, GPT4 and GPT4-CGLLM predictions.