Gate · Benchmark reference

CBLUE

Classic multi-task Chinese medical NLP checkup benchmark (entity, relation, classification, some generation). A pretraining checkup table — not a release gate.

Medical Gate private evaluation uses frozen private item sets, scoring rules and sample strata agreed in the contract. Public benchmarks are method reference only and do not become acceptance criteria unless both parties freeze them in writing.

Book Medical Gate Gate catalog Medical Model Eval

Contracting entity: Beijing Langhui Technology Co., Ltd. (brand LumeSage). Not a medical-device promotion; no clinical efficacy claims.