CI-Bench: Benchmarking Contextual Integrity of AI Assistants on Synthetic Data

Open in new window