From 684934e042cabacf736f4f57eb5b952f44e4d1be Mon Sep 17 00:00:00 2001 From: kiyo9w Date: Tue, 29 Sep 2026 04:31:59 +0900 Subject: [PATCH] docs(decision): re-measure E-3 and fix the labels README's dead reference E-3 with the calibrated lines: 66% at 95% precision with the text lane, 55% at 81% without, 2 false accepts against 8. The lane adds 11 pp of coverage while raising precision 14 pp, so the answer stands and the margin is wider than the audit's. The labels README pointed at evidence/image6_v2.json, which was a byte-identical duplicate of eval/labels/image6.json and is now removed; the README says so instead of describing a copy that no longer exists. --- .../2026-09-28-recognition-architecture/DECISION.md | 5 +++++ recognization/eval/labels/README.md | 9 +++++---- 2 files changed, 10 insertions(+), 4 deletions(-) diff --git a/recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md b/recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md index edc0916..56bf196 100644 --- a/recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md +++ b/recognization/docs/research/2026-09-28-recognition-architecture/DECISION.md @@ -815,6 +815,11 @@ against 20). Live: 62% at 93% against 56% at 84% (3 against 7). The lane adds 16 pp and 6 pp of coverage while *raising* precision, against E-3's +5 pp bar. It is not a veto-only lane. +Re-measured 2026-09-29 with the calibrated accept lines: 66% at 95% precision +with text, 55% at 81% without (2 false accepts against 8). The lane still adds +11 pp of coverage while raising precision by 14 pp, so the answer is unchanged +and the margin is wider than the audit's. + **E-2 answered.** `LocalLane` is RootSIFT-only; ORB and BRISK survive only in the `matcher` baseline lane, so there is nothing to drop from the shipped path. diff --git a/recognization/eval/labels/README.md b/recognization/eval/labels/README.md index 7188af0..a2bee5d 100644 --- a/recognization/eval/labels/README.md +++ b/recognization/eval/labels/README.md @@ -30,9 +30,10 @@ complete; every scorer drops them before computing a rate. The evaluators read in-catalog across 12 SKUs, 45 `unknown`. 26 labels changed from v1, and three more changed on 2026-09-29: facings #46-48 were `unknown` with the note "possibly what stock.csv calls l12", and the l12 reconciliation settled it, so -they now carry `sku_l12`. It is a copy of -`docs/research/2026-09-28-recognition-architecture/evidence/image6_v2.json`; -edit the evidence copy and re-copy, so the two cannot drift. +they now carry `sku_l12`. This file is the single copy: the evidence directory +used to hold a duplicate (`image6_v2.json`) and it was removed on 2026-09-29 +after `cmp` showed the two byte-identical, because a copy that nothing reads is +a copy that drifts. ## Schema check @@ -50,4 +51,4 @@ are absent rather than guessed. Tier-2 labels must add both. `confidence` is the labeler's own certainty for that row; `labeler` names the person or pass that produced it, because Tier 2 is double-labelled and a disagreement has to be traceable to a pass. Tier 1 -(`image6_v2.json`) is dev-only and its numbers never ship, so it stays as it is. +(`image6.json`) is dev-only and its numbers never ship, so it stays as it is.