sdil.git - Unnamed repository; edit this file 'description' to name the repository.

Age	Commit message (Collapse)	Author
11 days	figures: make audited PDFs deterministic	YurenHao0426

11 days	docs: freeze native baseline reproductions	YurenHao0426

11 days	results: close normalized C2 recovery	YurenHao0426

11 days	experiments: freeze normalized C2 recovery	YurenHao0426

11 days	sdil: normalize causal vectorizer regression	YurenHao0426

11 days	theory: prove estimator and innovation regimes	YurenHao0426

11 days	results: close feedback-first C2 recovery	YurenHao0426

11 days	experiments: freeze feedback-first C2 screen	YurenHao0426

11 days	sdil: add feedback-first apical calibration	YurenHao0426

11 days	results: close calibration quality recovery	YurenHao0426

11 days	experiments: freeze C2 calibration quality recovery	YurenHao0426

11 days	results: confirm C2 feedback amortization bottleneck	YurenHao0426

11 days	experiments: freeze direct perturbation C2 diagnosis	YurenHao0426

11 days	baseline: add unamortized node perturbation	YurenHao0426

11 days	results: close context-vectorizer C2 recovery	YurenHao0426

11 days	results: retain failed context-vectorizer recovery	YurenHao0426

11 days	experiments: freeze context-vectorizer C2 recovery	YurenHao0426

11 days	fix: bound state-conditioned feedback features	YurenHao0426

11 days	method: condition apical vectorizers on neural state	YurenHao0426

11 days	results: close failed C2 local panel	YurenHao0426

11 days	results: retain failed useful-depth local panel	YurenHao0426

11 days	experiments: freeze useful-depth local-rule panel	YurenHao0426

11 days	results: mark useful-depth BP screen passed	YurenHao0426

11 days	results: establish BP useful-depth candidate	YurenHao0426

11 days	experiments: freeze useful-depth BP screen	YurenHao0426

11 days	experiments: audit residual block usage with lesions	YurenHao0426

11 days	results: close failed C1 recovery branch	YurenHao0426

11 days	results: retain failed FashionMNIST C1 recovery	YurenHao0426

11 days	experiments: freeze FashionMNIST C1 recovery	YurenHao0426

11 days	results: record frozen C1 near-miss	YurenHao0426

11 days	results: confirm mixed-traffic innovation across seeds	YurenHao0426

11 days	experiments: freeze mixed-traffic innovation confirmation	YurenHao0426

11 days	experiments: isolate final diagnostics from held-out data	YurenHao0426

11 days	results: mark amortized calibration gate passed	YurenHao0426

11 days	results: confirm low-query calibration across seeds	YurenHao0426

11 days	experiments: freeze five-seed low-query confirmation	YurenHao0426

11 days	experiments: log peak memory and total training work	YurenHao0426

11 days	results: freeze single-direction calibration	YurenHao0426

11 days	results: identify 16x lower-query calibration	YurenHao0426

11 days	experiments: predeclare low-query calibration screen	YurenHao0426

11 days	experiments: isolate apical traffic seeds	YurenHao0426

11 days	results: freeze predictor timescale choice	YurenHao0426

11 days	docs: correct SDIL novelty boundary	YurenHao0426

11 days	results: map neutral predictor timescales	YurenHao0426

11 days	experiments: sweep neutral predictor timescales	YurenHao0426

11 days	experiments: audit calibration cost and isolate timing	YurenHao0426

11 days	docs: remove unsupported synthetic depth guarantees	YurenHao0426

11 days	results: record failed useful-depth BP screens	YurenHao0426

11 days	experiments: add useful-depth validation screen	YurenHao0426

11 days	experiments: add synthetic validation isolation	YurenHao0426