From 167312abe9798d691bd059a36f1d615903d355de Mon Sep 17 00:00:00 2001 From: Yuren Hao Date: Tue, 28 Jul 2026 17:43:11 -0500 Subject: =?UTF-8?q?RESULT=2067:=20=E4=BB=AA=E5=99=A8=E9=87=8D=E6=A0=87?= =?UTF-8?q?=E5=AE=9A(0.0685=3D33%harness=20confound+67%=E7=89=A9=E7=90=860?= =?UTF-8?q?.0462);=20=E7=94=B5=E6=B1=A02=E5=85=A8=E7=81=AD(headmix=200%=3D?= =?UTF-8?q?=E8=8A=82=E6=B5=81=E6=98=AF=E7=97=87=E7=8A=B6,R66=E5=9B=A0?= =?UTF-8?q?=E6=9E=9C=E9=93=BE=E6=92=A4=E5=9B=9E);=20=E5=90=8C=E5=A3=B3?= =?UTF-8?q?=E5=AF=B9=E7=85=A7=E7=BA=AA=E5=BE=8B=E5=85=A5=E5=86=8C;=20?= =?UTF-8?q?=E7=94=B5=E6=B1=A03=E5=AE=9A=E4=BD=8D(=E5=BF=85=E5=88=86?= =?UTF-8?q?=E4=BF=A1=E5=8F=B7)=E5=9C=A8=E9=A3=9E?= 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 | 14 ++++++++++++++ 1 file changed, 14 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 5c4b25c..6c11cf2 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -568,6 +568,20 @@ direction), not training-under-fault; wave-2 = co-training with faults injected +### 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. +Battery 2: leak_headmix (head on TRUE BP grads, config-verified) closes 0%; head_lr_mult +1.67/2.5 closes -1%/-3%. R66's causal chain RETRACTED (the 60% throttle measurement stands, +but fixing it buys nothing — the head lags because of something upstream, or harmlessly). +INSTRUMENT CONTROL leak_allbp (EP harness + ALL-BP grads, same data order): 3.5206 vs +casc_bp_train 3.4982 vs EP arms 3.5668 -> the 0.0685 raw leak = 0.0224 harness/data-order +artifact (33%) + 0.0462 gradient-physics (67%). Battery 1-2 conclusions survive on the +recalibrated denominator: five EP configs all at 3.5668 = 0% of 0.0462 closed. Lesson banked: +cross-harness comparisons REQUIRE the same-harness control arm first (amp-probe lesson, trainer +edition). Battery-arm discipline: reference configs must pin every factor explicitly. +BATTERY 3 (localization, must-split by construction): bpmix blocks:6-11 / blocks:0-5 / attn. + ### RESULT 66 (2026-07-28): THE HEAD IS THROTTLED TO 60% — user's equivalent-LR hypothesis CONFIRMED and localized to ONE matrix; symmetrization/precision/depth all irrelevant to it. Displacement micro-probe (600 steps from s250000, ||dW||/||W|| per family): every family -- cgit v1.2.3