NoveltyScope, described in a new arXiv preprint, is a prototype research assistant designed to assess how novel a scientific paper is. It works by retrieving prior work related to the paper and then comparing the paper against that body of literature.

The system uses a multi-agent approach to generate a structured report and a five-level novelty score. It identifies the paper's core task and extracts at most two author-claimed contributions to focus the assessment. As a prototype, it is an early step toward automating novelty evaluation, though the abstract does not detail the specific agents or scoring criteria.