From f75e1f0877c7d1b57299a68507d30eeb16bf3b4d Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Sat, 1 Aug 2026 17:02:43 -0500 Subject: Correction, same session: the width statistic does not survive its control The replacement gate proposed this morning is refuted by a control run this afternoon. Independent noise lowers the correlation to 0.828 without narrowing the underlying signal and reaches the same measured width as a field whose captions omit one factor -- 15.7 against 15.0 -- with recovery at 94.7% and 5.6%. A second instrument built specifically to fix that, counting canonical directions that generalise to held-out scenes, fails the same way and rates Visual Genome highest of the three failing fields. Established: the correlation does not govern recovery, in both directions. Corpus overlap does control it -- the caption-suppression ladder is monotone from 96.4% to 0.0%. Not established: any statistic that predicts recovery cheaply. Documents corrected accordingly rather than quietly rephrased; the refuted claim stood for four hours and is recorded as such. Co-Authored-By: Claude --- CANDIDATE_REGISTER.md | 26 +++++++++++++++----------- 1 file changed, 15 insertions(+), 11 deletions(-) (limited to 'CANDIDATE_REGISTER.md') diff --git a/CANDIDATE_REGISTER.md b/CANDIDATE_REGISTER.md index e4aa235..bcfa3a0 100644 --- a/CANDIDATE_REGISTER.md +++ b/CANDIDATE_REGISTER.md @@ -5,24 +5,28 @@ The register used to be ordered by how much each candidate was expected to raise the field correlation, because correlation was the go/no-go statistic. That statistic has been shown not to govern recovery: a synthetic field -truncated to rank 8 correlates at 0.906 and recovers 13%, while the same field -at rank 16 recovers 96%. **The target is now a pair — the correlation and the -width of the shared spectrum — and candidates are ordered by which one they -move.** +truncated to rank 8 correlates at 0.906 and recovers 13%, while independent +noise that drops the correlation to 0.743 leaves recovery at 71%. Two +replacement statistics were then proposed and both failed their own controls +(see `RANK_RESULTS.md`), so **there is currently no cheap predictor and +candidates have to be scored by running the search.** Rows below keep their +width measurements because they were taken, not because they predict anything. Status is one of: **done** (measured, number recorded), **running**, **queued** (specified, not started), **open** (idea, not specified), **closed** (measured and eliminated). -Current position: correlation 0.731, shared directions 15, against a -recovering synthetic reference at 0.928 and 26. +Current position: correlation 0.731 and recovery 0.0000 against 0.0039 +chance, versus a synthetic reference at 0.928 and 95.8%. -## A. Widen the shared spectrum +## A. Widen what the two corpora share -The binding constraint, and the one only identified today. Both modalities are -individually rich — vision effective rank 40, text 48 — and agree on 15 -directions. Candidates here are judged on the shared count, not the -correlation. +The live constraint, though the quantity that expresses it is not yet +identified. What is established by intervention: suppressing one factor from +the synthetic captions takes recovery from 96.4% to 5.6% while the correlation +stays at 0.830, and five separate encoder-side interventions move neither +recovery nor any width measure. Candidates here are judged by whether they +change what the two corpora are about. | Candidate | Cost | Status | |---|---|---| -- cgit v1.2.3