Backdoor attacks are a serious threat to AI security. In this competition, you will explore backdoor detection in code summarization datasets.
You are given a dataset (`train.csv`) containing Python functions and their corresponding natural language docstrings. Most samples are clean, but a small fraction (~1%) are poisoned with hidden triggers that can mislead models.
Your task is to build a classifier that predicts whether a given (code, docstring) pair is clean or poisoned.
This assignment is part of CSC591-028: Generative AI for Software Engineering at NC State University.