The abstract introduces RATIO, a benchmark for retrieval across typed ideation operations in scientific literature. It positions retrieved papers as a source of inspiration for both human and AI researchers, suggesting that the way prior work informs new ideas is not uniform.

According to the source, inspiration can take different forms: prior work may directly suggest how to address a problem, or it may surface directions at different levels of abstraction. RATIO appears designed to capture and evaluate these distinct retrieval needs.

Because the abstract is truncated, the source does not provide details on the benchmark's construction, evaluation methodology, or results. The description is limited to the motivation and the typology of ideation operations.