SCOUT: Cascaded Visual Document Retrieval Cuts Latency Up to 80.7x While Beating SOTA by 7.8% NDCG
_reachsumit · x · 2026-10-15
A new arXiv paper proposes SCOUT, a cascaded retrieval framework for visually rich documents, tackling the query-time cost of vision-language retrievers that grows with corpus size.
Method highlights:
- Two-track cascade: a visual track uses a lightweight parent retriever to filter the corpus into a compact candidate pool, then applies a stronger multi-vector reranker only on candidates; a semantic track retrieves over hierarchical page/document summaries generated offline, providing a signal complementary to visual matching
- Cross-modal query refinement: at fusion, semantic scores adjust the visual query representation before final ranking
Results: across 4 major benchmarks vs 6 baselines, query-time latency improves up to 80.7x (28.8x average) over the best baseline while exceeding state-of-the-art retrieval by up to 7.8% NDCG.
More from Research
- kalomaze argues ensembled synthetic data can cover data-scarce domains — kalomaze · 2026-10-15
- Axiom claims proof of generalized quasi-Riemann hypothesis for all Dedekind zeta functions — basedjensen · 2026-10-15
- TorchLean expands into a verified tensor programming framework with einsum and GPU execution in Lean — burny_tech · 2026-10-15
- StreamErase hits 26 FPS real-time video object erasing on a single GPU, 200x faster than prior methods — Zhuobai Dong · 2026-10-15
- Huawei's SwarmRSI: evolving agent swarms achieve recursive self-improvement in science — Qianru Meng · 2026-10-15
- Glance Forcing cuts video distillation trajectory cost from thousands of GPU-hours to one — Zhuobai Dong · 2026-10-15