GLIE Paper Compresses Visual Document Retrieval Storage
A KAUST arXiv preprint, GLIE, tackles the storage burden of late-interaction visual document retrieval by reconstructing full embeddings from only about 4 vectors per page, drastically compressing storage.
2026-09-11 ~ 2026-09-11 · 3 related posts
- GLIE regenerates full page embeddings from 4 vectors, slashing visual document retrieval storage — _reachsumit · 2026-09-11
- GLIE: Late-Interaction Embeddings Have Only ~5 Degrees of Freedom, Enabling Massive Storage Savings — lateinteraction · 2026-09-11
- KAUST Proposes Generative Late-Interaction Embeddings to Compress Visual Document Retrieval Vectors — KAUST · 2026-09-11