Product-search benchmark with Exact, Substitute, Complement, and Irrelevant relevance labels for query-product pairs.
Retail Search
ARCH-0.5 · cross-domain core
Phase 3 - Retail Relevance (Amazon ESCI)
Product search over 1.2M Amazon products with structured fields. The BGE hybrid reaches parity with the published baseline (0.8468 vs 0.8503) and understands intent keyword search cannot.
Livestatus
Amazon ESCI (1.2M products)dataset
Official hybrid nDCG 0.8468 (baseline 0.8503)metrics
Live retail search
Search 1.2M Amazon products. Fast keyword ranking by default; the BGE hybrid (semantic) is opt-in and understands intent keyword search cannot.
semantic searchBGE hybrid: intent understanding
field-level explainwhy a product ranked
/api/esci/evaluationbaselines + hybrid parity
/api/esci/metacorpus + architecture