We summarize our schedule into the following milestones with a corresponding timeline:
Milestone 1: To do literature review, demonstrate problem, and show existing methods including making unlearnable texts examples cannot work in the setting of RAG systems. To be completed in October, 2024.
Milestone 2: To explore plausible solutions in the setting with constraints of using one specific RAG system (LLM + Retriever) under a white-box scenario. To be completed in November, 2024 – December, 2024.
Milestone 3: To further extend the solutions to the black-box scenario with multiple plausible LLMs. To be completed in January, 2025 – March, 2025.
Milestone 4: To quantitatively analyze our solution. To be completed in April, 2025.