Scientific literature is vast and contains within it as yet unnoticed connections. Literature-based discovery is an attempt to bring them to light.

We evaluate an automated approach for catching hallucinations in paper abstracts, aiming for consistently more trustworthy results.