RerankIndex lands in kANNolo 0.12: two-stage search framework beats HNSW on QPS
CShorten30 · x · 2026-10-14
The tusKANNy team released RerankIndex in kANNolo 0.12 (Python & Rust), a general abstraction framework for two-stage (retrieve-then-rerank) vector search, aiming to simplify a fragmented retrieval landscape.
Benchmarks show it consistently outperforms HNSW in single-thread QPS across Accuracy@10 operating points on NQ (SPLADE CoCondenser, 2.7M vectors) and MS MARCO (SPLADE-v3, 8.8M vectors), with gaps exceeding 1.5x on MS MARCO.
More from Research
- Sakana AI proposes MASS: multi-agent self-supervision for recursive self-improvement — paulabartabajo_ · 2026-10-14
- UNITAS: a 1.7B 3D-native world action model hits 99.8% on LIBERO and 85% on real tasks — CUHKSZ · 2026-10-14
- Joint training of retrieval representations lifts all boats, BGE-m3 and NeoMME findings — antoine_chaffin · 2026-10-14
- Crowd of agents push integer multiplication bounds with 'slop-search', kappa records falling every few minutes — iskander · 2026-10-14
- New paper predicts alignment generalization from value representations, builds first LLM value taxonomy — VeredShwartz · 2026-10-14
- 'Two-way RL' for task-level model compression: building wafer-thin hyper-specialized models — tokenbender · 2026-10-14