2 Commits

Author	SHA1	Message	Date
Igor Lins e Silva	7e4db33061	fix: resolve ruff lint errors in benchmark suite ... Remove unused imports (shutil, string, datetime, os, yaml, time, SCALE_CONFIGS) and unused variable assignments in timing-only calls.	2026-04-08 05:10:39 -03:00
Igor Lins e Silva	e8017ca2ec	bench: add per-room recall threshold test ... Concentrates all drawers into a single wing+room to isolate the embedding model's retrieval limit independent of palace filtering. Confirms recall degrades to ~0.4-0.5 at 5K drawers per room even with wing+room filters applied — the spatial structure helps by keeping buckets small, but can't fix the underlying embedding ceiling.	2026-04-08 05:06:31 -03:00