zhliu/ArxivMIA

Dataset Card for ArxivMIA To evaluate various pre-training data detection methods in a more challenging scenario, we introduce ArxivMIA, a new benchmark comprising abstracts from the fields of Computer Science (CS) and Mathematics (Math) sourced from Arxiv. Repository: https://github.com/zhliu0106/probing-lm-data Paper: Probing Language Models for Pre-training Data Detection