Multimodal rerankers now have HF Spaces

CShorten30 · x · 2026-07-20

The authors have put their new multimodal rerankers behind a set of Hugging Face Spaces so people can try them without deploying anything locally.

The screenshot shows a multimodal listwise reranking playground where you upload four document page images, choose a model size, enter a query, and get a ranked order back. It’s positioned as an easy way to experiment with their rerankers on text-plus-image retrieval tasks.

Original post →

More from Multimodal

Multimodal channel →