data/ground-truth-retrieval.csv

video_id,question
zjkBMFhNj_g,What is prompt injection and how does it work as an attack on language models?
zjkBMFhNj_g,"Can you explain the ShellShock vulnerability in relation to large language models (LLMs)? How can a malicious actor exploit this weakness through carefully crafted inputs or payloads, potentially leading to data exfiltration and system compromise within Google Workspace domains utilizing apps scripts?"
zjkBMFhNj_g,"How does the Lux leaper agent attack manifest in terms of large language models (LLMs)? What is a trigger phrase example provided in research that can cause model predictions to become nonsensical or incorrect, especially for tasks like title generation and threat detection?"