jspace-nanogpt.git - J-space on nanoGPT: Jacobian Lens replication & critique. J-lens norms anti-correlate with token frequency (r ~ -0.61 to -0.69); frequency-matched synthetic pair + causal loss-reweighting tests.

Age	Commit message (Collapse)	Author
2 days	Blog: clarify J-lens frequency findingsHEAD main	CaptainJack2491

2 days	Visualize J-lens frequency findings	CaptainJack2491

4 days	at-scale W_U row-norm check (GPT-2): geometric confound generalizes, ↵	Void Agent
	r(log10)≈-0.45/-0.49; blog draft: drop 'Actually', log-freq lead, probe transparency, norm-vs-readout concession
7 days	Short version: clean-boundary control is resolved (~1.3x), not pending	Void Agent

7 days	Polish per review: results.md GPT-2 -0.18 demoted to do-not-cite footnote; ↵	Void Agent
	blog section 10 adds W_U-at-init scale prediction (one-line GPT-2 unembedding check)
7 days	Blog revision per reviewer feedback: de-emphasize reviewer name-drops, ↵	Void Agent
	promote LN technical note, willkn as Related-work, table footnote, section 8 power reframe (~13 seeds/arm needed); add W_U-at-init finding (learned geometry, r~0 at init vs -0.61 trained)
7 days	Clean-boundary synthetic-pair control: ratio 1.311 (CI [1.258,1.402]) vs ↵	Void Agent
	1.471 — corruption inflated by ~12%, structure signal survives at 0% word-slicing; blog section 7 + results.md resolved
7 days	Blog (Claude's review): add 'short version' TL;DR up top for wider audience; ↵	Void Agent
	promote W_U decomposition to its own flagged subsection as the novel mechanistic contribution
7 days	Blog: fill in public repo URL (git.jayrup.me/c/jspace-nanogpt.git)	Void Agent

7 days	Robustness confirmed: freq correlation identical under Anthropic's exact ↵	Void Agent
	estimator (skip16+source_mean), max delta 0.008 all layers
7 days	Fix in-word metric: strict (both-sides, 58% random / 0% clean) vs touches ↵	Void Agent
	(95%); blog/results caveat corrected to the strict discriminator
7 days	Loss-reweighting results: q/control 0.944 (CI crosses 1), q/ctrl_random ↵	Void Agent
	0.985 — small non-significant causal demotion; freq corr invariant (r~-0.66) across all 9 models; blog section 8 + results.md section 3 filled honestly
7 days	Add willkn GreaterWrong corroboration: independent GPT-2-medium finding that ↵	Void Agent
	raw J-lens misweights structural (high-frequency) tokens + full-rank Jacobian; shrinkage J+lambda I as untested next experiment
7 days	Verify against Anthropic's official jacobian-lens repo: estimator match, ↵	Void Agent
	zero frequency in their code+data, add --skip_first/--source_mean mirror mode
7 days	Address Luna repo review: dynamic repo root (no hardcoded cwd), J-lens ↵	Void Agent
	correctness test (last-layer identity), auditable fact-check, mechanism narrative fix, softened conditional-predictability claim, README data-prep + artifacts note
7 days	Address Gemini+Opus repo reviews: W_U decomposition, Spearman+p-values, ↵	Void Agent
	synthetic-pair CIs + corruption caveat, accurate fact-check, softened burden-of-proof, LayerNorm note, reviews caveat
7 days	Blog: fact-check section — 3 independent checks confirm Anthropic has no ↵	Void Agent
	frequency control (incl. template-lens high-frequency note); preempt 'they have controls' strawman
7 days	Docs: Feynman-style blog draft, MIT license, requirements, results.md, ↵	Void Agent
	README rewrite with repro steps; reviews -> docs/reviews