From 0f0bce37ddbc485768181e3f4d8bf93ba660c5db Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Wed, 8 Apr 2026 09:28:35 -0500 Subject: =?UTF-8?q?Round=2040=20=C2=A74=20update:=20cos-vs-acc=203-part=20?= =?UTF-8?q?proposition=20(Observation=20/=20Inference=20/=20Mechanism=20hy?= =?UTF-8?q?pothesis)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Codex round 40 recommended turning the Mode 2 dissociation from empirical curiosity into a methodological claim via a 3-part proposition: 1. Observation: CB+pen and DFA+pen reach same acc despite 4x deep-cos gap; SB+pen best acc with intermediate cos 2. Inference: layerwise BP-cosine is NECESSARY to rule out grossly wrong credit signals but NOT SUFFICIENT to certify usable credit for depth 3. Mechanism hypothesis: usefulness depends on whether local updates induce coordinated forward-state change across blocks, not just angular agreement with BP Method framing (codex-approved safer versions): - CB = 'gradient-direction surrogate' (high angular agreement, low functional credit) - SB = 'state-level downstream teaching signal' (lower angular agreement, higher functional credit) - Explicitly framed as HYPOTHESIS not theorem Main content still 9 pages exactly (within E&D limit). Co-Authored-By: Claude Opus 4.6 (1M context) --- paper/main.pdf | Bin 477256 -> 478989 bytes 1 file changed, 0 insertions(+), 0 deletions(-) (limited to 'paper/main.pdf') diff --git a/paper/main.pdf b/paper/main.pdf index c2510e5..dcb3fa4 100644 Binary files a/paper/main.pdf and b/paper/main.pdf differ -- cgit v1.2.3