From 71096d092978f7e2b7466b4a98dbc06ef67aabbd Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Sat, 18 Jul 2026 23:04:14 -0500 Subject: =?UTF-8?q?RESULT=2049:=20=E5=9B=9E=E8=B7=AF=E8=A7=A3=E5=89=96=3D2?= =?UTF-8?q?3=E6=AE=B5=E4=B9=98=E7=A7=AF=E8=BF=91=E4=B8=B4=E7=95=8C(?= =?UTF-8?q?=E4=B8=8B=E8=A1=8C=C3=9772/=E4=B8=8A=E8=A1=8C=C3=9713k/?= =?UTF-8?q?=E9=A1=B6=E8=AF=BB1e-6),plain=E5=8F=AA=E7=83=ADb8=E4=B8=80?= =?UTF-8?q?=E6=AE=B5(h6=20logit=2094=20vs=20cent=2075)=E8=BE=B9=E9=99=85?= =?UTF-8?q?=E8=B6=8A=E7=BA=BF,=E7=81=BE=E5=8F=98=3D=E7=A1=AC=E5=BA=95?= =?UTF-8?q?=E5=93=8D=E5=BA=94=E6=89=93=E4=BC=A4b1(=E5=85=A5=E5=8F=A3vjp=20?= =?UTF-8?q?1.3=E2=86=9214.9);=20batch=E5=99=AA=E5=A3=B0=E5=AF=B9=E7=85=A70?= =?UTF-8?q?.39/0.45=3DQ1=E5=B0=81=E5=8F=A3;=20--logit=5Fknee=2080=E5=AE=9E?= =?UTF-8?q?=E8=A3=85+knee=E8=87=82=E6=8E=92c190=E5=90=8E?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn --- docs/campaign/CASCADE_ABLATION_PLAN.md | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) (limited to 'docs/campaign/CASCADE_ABLATION_PLAN.md') diff --git a/docs/campaign/CASCADE_ABLATION_PLAN.md b/docs/campaign/CASCADE_ABLATION_PLAN.md index 268ab91..f4471d6 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -568,6 +568,28 @@ direction), not training-under-fault; wave-2 = co-training with faults injected +### RESULT 49 (2026-07-18): CYCLE-GAIN DECOMPOSITION — the explosion anatomized (user +directive). The loop is a ~23-stage product sitting near-critical by construction; plain heats +ONE stage (b8's sharp head h6, logit 94 vs cent 75) until a MARGINAL crossing; the catastrophe +was the governor's hard-bottom RESPONSE, which damaged b1 (vjp-entry gain 1.3 -> 14.9). +probe_cycle (mode-aligned per-stage gains, sweeps 15/16 diffs, beta=3e-3): + structure (BOTH lineages): down-chain vjp product ~50-72x, up-chain rebuild ~13-15k x, + top read gH ~1e-6 (beta*NBT*H_top) -> round-trip O(1). Biggest single amplifier = b8: + 2.9 (vjp) x 3.8 (fwd) ~ 11x round-trip; b7 next. Same map as R44/R47. + plain vs cent @195k: same structure ~1.2-1.4x hotter, concentrated at the b8 stage + (2.92 vs 2.50) and its carrier head b8-h6 (share 0.21, max|logit| 94 vs cent 75, ent 0.4); + healthy runs CARRY 50-57-logit heads fine (cent b10-h7=57, b11-h0=51) -> threshold ~80+. + plain @210k (sick): gV entry-stage 0<-1 EXPLODED 1.31 -> 14.91 (b1 sigma(J) 16->34->44 in + statej) = the noise-walk DAMAGE, not the original cause. Pre-collapse growth was modest + (res0 +55% over 40k steps); G's 300x conflates marginal crossing with post-crossing damage. +BP-batch-noise control: per-batch BP u1 vs mean-BP u1 overlap 0.388/0.446 == EP arms' 0.37-0.41 +-> the "shared noise" is minibatch sampling, identical for BP; Q1 sealed. +FIX ARM ARMED: --logit_knee 80 (piecewise: identity below 80, slope 0.2 above; healthy heads +untouched incl cent's own carriers; cuts the runaway head's switching Jacobian 5x). +fw72m_plain_knee = exact plain_nf replay + knee, queued behind c190 (smoke then DDP launch). +Also on the menu (user decision): est_auto = centered-on-demand only when rho_ema near cap +(pays 1.39x on ~10-20% of steps, avg ~1.05x) — mechanism-blind but c190-proven lever. + ### RESULT 48c (2026-07-18): WHY the digital floor can't be BBP — the THETA-READ HAS NO 1/beta CHANNEL, structurally. The BBP transition lives exactly where readout is DIFFERENTIAL (analog). probe_bbp3 (6-bit STOCHASTIC rounding, fresh draw per estimate; frozen = one draw for free+ -- cgit v1.2.3