Skip to content

Latest commit

 

History

History
10 lines (7 loc) · 501 Bytes

File metadata and controls

10 lines (7 loc) · 501 Bytes

SciMON: Scientific Inspiration Machines Optimized for Novelty

Paper: https://arxiv.org/pdf/2305.14259

SciMON uses three graphs (citation, semantic and KG) to retrieve inspiration terms that are used by the LLM to generate novel scientific ideas.

This module is a reproduction of this method using the same principles but different methods. In particular, SciMON uses a pipeline of models to extract the key terms and context from the paper abstracts, whereas we use GPT prompting to do the same.