Tag
This paper proposes a categorical approach to generate falsifiable research ideas by modeling papers as small categories and using a functor-preservation gate to filter cross-domain analogies, improving over traditional LLM-based methods.