From c67f91ba8f1eb70c64897641bb88cee4af249ed2 Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Tue, 28 Jul 2026 19:44:35 -0500 Subject: =?UTF-8?q?RESULT=2068:=20=E6=B3=84=E6=BC=8F=E5=AE=9A=E4=BD=8D?= =?UTF-8?q?=E6=94=B6=E5=AE=98=20=E2=80=94=20100%=E4=BD=8F=E9=A1=B6?= =?UTF-8?q?=E5=8D=8Ablock=E6=A2=AF=E5=BA=A6,attn=E5=8D=A02/3,=E4=B8=8ER44?= =?UTF-8?q?=E5=90=8C=E5=9B=BE(=E4=B8=80=E4=B8=AA=E7=BB=93=E6=9E=84?= =?UTF-8?q?=E4=B8=A4=E4=B8=AA=E5=B0=BA=E5=BA=A6,=E5=AE=BD=E5=BA=A6=C3=976)?= =?UTF-8?q?;=200.999-cos=E6=82=96=E8=AE=BA=E6=9C=89=E4=BA=86=E5=9C=B0?= =?UTF-8?q?=E5=9D=80=3D=E9=A1=B6=E9=83=A8=E5=B0=96=E6=B3=A8=E6=84=8F?= =?UTF-8?q?=E5=8A=9BEP=E8=AF=BB=E6=95=B0~1e-5/=E6=AD=A5=E7=B3=BB=E7=BB=9F?= =?UTF-8?q?=E8=B5=A4=E5=AD=97;=20=E7=94=B5=E6=B1=A04(=CE=B2=E5=89=82?= =?UTF-8?q?=E9=87=8F)=E5=9C=A8=E9=A3=9E=E5=88=A4=E4=BD=8D=E7=A7=BB?= =?UTF-8?q?=E9=80=9A=E9=81=93?= 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 | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'docs/campaign') diff --git a/docs/campaign/CASCADE_ABLATION_PLAN.md b/docs/campaign/CASCADE_ABLATION_PLAN.md index 6c11cf2..52c266e 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -568,6 +568,21 @@ direction), not training-under-fault; wave-2 = co-training with faults injected +### RESULT 68 (2026-07-28): LEAK LOCALIZED — 100% in the TOP-HALF block gradients, ~2/3 +attention; the SAME map as R44 at 72M, 6x stronger at width. One structure, two scales. +Battery 3 (guaranteed-signal partition, bars allbp 3.5206 / EP 3.5668): mixtop closes 100% +(tail = allbp bar exactly), mixbot 0% (tail = EP bar exactly), mixattn 67%. The R44 endgame +decomposition (77/0/41-37) reproduces on the leak instrument at 135M. The 0.999-cos paradox +now has an address: top-half sharp-attention EP reads carry a ~1e-5/step systematic CE deficit, +invisible to every per-step metric, killed only by BP substitution. +Suspect state after batteries 1-3: NOT odd-order bias (bsign/centered), NOT read noise +(centered/fp32), NOT relax depth (K8), NOT the head (headmix/hlr), NOT bottom transmission +(mixbot 0%). Remaining channels for a top-local, symmetric-in-beta(?), BP-substitutable +deficit: anchor-displacement curvature interaction at sharp softmax, or d-signal structure at +the top read. BATTERY 4 (beta dose 1.5e-3/6e-3, same harness): leak ~ beta -> displacement +channel confirmed -> treatment = top-local beta reduction (one flag); leak flat in beta -> +displacement out, next = top-local read variants. + ### RESULT 67 (2026-07-28): INSTRUMENT RECALIBRATED (1/3 harness confound, 2/3 physics) + BATTERY 2 ALL-NEGATIVE — the head-throttle is a SYMPTOM, not the cause; the 0.046 physics leak survives every single-component surgery. Localization battery (guaranteed-signal) in flight. -- cgit v1.2.3