SOURCE-LINKED INTELLIGENCE

Spatial Matryoshka Training for Multi-Granularity Visual Document Retrieval

arXiv · AI, language, vision and robotics · article · Aug 30, 2026 · UTC

Multi-modal late-interaction retrievers achieve strong retrieval on visually rich documents by representing each page as per patch embeddings and matching at the token level. However, this approach incurs high storage costs. Existing compression methods typically fix a single compression level at indexing time, limiting flexibility. We present ColSNAP (Spatial Nested Average Pooling)1, a training method that generates a nested hierarchy of compression levels directly from a backbone's patch grid. By spatially pooling patch embeddings into pro- gressively coarser tiers and training all tiers si

Read original source ↗ Open in workspace

recordType

paper

region

Global