From d344f7fa4b8a6b907d364d1900348ca8e4b709ce Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Wed, 22 Jul 2026 04:51:49 -0500 Subject: results: confirm C2 feedback amortization bottleneck --- NOVELTY.md | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) (limited to 'NOVELTY.md') diff --git a/NOVELTY.md b/NOVELTY.md index 3755595..c044927 100644 --- a/NOVELTY.md +++ b/NOVELTY.md @@ -76,11 +76,16 @@ identifies the traffic relationship without observing the teaching component. - The inherited learned-feedback backbone preserves performance with depth better than fixed DFA in the audited flattened-CIFAR MLP setting. -- Controlled pilots show that per-cell residualization can protect learning from soma-coupled and - endogenous top-down traffic; the latter remains a validation-only, single-seed result until C1 is - completed. -- Simultaneous batched perturbations make the inherited calibration practical in this codebase; - a hardware-independent query/FLOP advantage is not yet established. +- Frozen controls show that per-cell residualization protects learning from soma-coupled traffic. + It mitigates but does not fully remove endogenous top-down traffic, so arbitrary contextual + innovation is not an allowed teaching-signal claim. +- Simultaneous batched perturbations have a frozen hardware-independent advantage: K1/every4 + retains 112.9% of the K16/every4 gain over DFA with 16x fewer logical loss queries, 11x less + calibration work, and 5.3x less total forward-equivalent work on CIFAR-10 d20/w64. +- Direct node perturbation solves the frozen useful-depth task while the amortized context + vectorizer does not. This identifies feedback amortization as the current engineering + bottleneck; it is not a novelty claim and the direct estimator's 68.4x work precludes presenting + it as the scalable algorithm. ## Claims currently forbidden -- cgit v1.2.3