A paper posted on arXiv addresses a shift in web-based visual search toward semantic intent—queries that focus on what is happening in an image rather than merely what it looks like. The authors propose PRISm, an importance-aware scene graph approach for semantic image retrieval.

The abstract begins to contrast this approach with global embeddings from vision-language models, but the sentence is cut off in the source. As a result, the digest can only report the stated motivation and the name of the proposed method; no experimental results or comparisons are available in the provided text.