You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Nov 21, 2026. It is now read-only.
Stale references after the 2026-10 retirement: DBBench page wording, dense id index switch, instrument sentences #448
After DECISIONS #196 (2026-10 and September retired, #429#431#432#434 merged) a sweep of README.md, bindings/python/docs, benchmarks/README.md, benchmarks/experiments/.md and the comments in benchmarks/experiments/.py found a few statements that no longer hold.
The docs still send readers to the ArcadeDB project page for benchmark tables (it now shows method only and points to DBBench, humem.ai#76), and still quote a measured outcome from retired 2026-10 rows: the docs index, benchmarks/README.md, the vector search, server mode and cross-model examples, and the overrides sentence in benchmarks/protocol.md.
The override registry and a page_check description still say the dense id index is off unless BENCH_DENSE_ID_INDEX=1, but every dense adapter builds it unconditionally (test_instrument_method.py holds that).
Sentences in CAMPAIGN.md, FAIRNESS.md, PROTOCOL.md and l1_tpc.py still name the 2026-10 instrument as the current one, or say September's rows stay as measured (they were removed). The gates they describe run on the published instrument, 2026-11.
Names that look stale but are current and stay: runs_paper.csv (the freeze), cap_rule_2610.json (the cap rule data the 2026-11 generator reads), PUBLISH_INSTRUMENT. The quantization PRODUCT text belongs to #446.
Activity