AIDA-Syn

AIDA-Syn is a synthetic dataset meant to be used as a diagnostic tool for contamination detection in LLMs. We provide smaller versions of the dataset (10, 50 and 100 documents) containing different alternatives than the full dataset. Do note that the document reference IRIs remain the same across the 4 provided datasets despite document mentions, mention spans and entities varying. The data is provided in NIF / TTL (NLP Interchange Format / Turtle) format and can be ingested using e.g. pynif or any knowledge graph database.